gh actions: upgrade to 1.2.11-1

Added: 增加IPv6支持

.gitignore

@@ -31,3 +31,4 @@ src/Makefile
 
 # other
 *.swp
+*.swo

debian/README.Debian

@@ -0,0 +1,3 @@
+
+this network service framework library extract from FastDFS
+

debian/changelog

@@ -0,0 +1,198 @@
+libserverframe (1.2.11-1) unstable; urgency=medium
+
+  * upgrade to 1.2.11-1
+
+ -- YuQing <384681@qq.com>  Sun, 23 Nov 2025 10:48:22 +0000
+
+libserverframe (1.2.11-1) unstable; urgency=medium
+
+  * upgrade to 1.2.11-1
+
+ -- YuQing <384681@qq.com>  Sun, 23 Nov 2025 10:00:56 +0000
+
+libserverframe (1.2.11-1) unstable; urgency=medium
+
+  * upgrade to 1.2.11-1
+
+ -- YuQing <384681@qq.com>  Sun, 23 Nov 2025 09:06:43 +0000
+
+libserverframe (1.2.8-1) unstable; urgency=medium
+
+  * upgrade to 1.2.8-1
+
+ -- YuQing <384681@qq.com>  Sat, 16 Aug 2025 16:32:03 +0000
+
+libserverframe (1.2.7-1) unstable; urgency=medium
+
+  * upgrade to 1.2.7-1
+
+ -- YuQing <384681@qq.com>  Sun, 06 Apr 2025 16:56:40 +0000
+
+libserverframe (1.2.5-1) unstable; urgency=medium
+
+  * upgrade to 1.2.5-1
+
+ -- YuQing <384681@qq.com>  Sun, 29 Sep 2024 15:24:39 +0000
+
+libserverframe (1.2.4-1) unstable; urgency=medium
+
+  * upgrade to 1.2.4-1
+
+ -- YuQing <384681@qq.com>  Sat, 15 Jun 2024 14:45:44 +0000
+
+libserverframe (1.2.3-1) unstable; urgency=medium
+
+  * upgrade to 1.2.3-1
+
+ -- YuQing <384681@qq.com>  Sun, 17 Mar 2024 15:11:04 +0000
+
+libserverframe (1.2.2-1) unstable; urgency=medium
+
+  * upgrade to 1.2.2-1
+
+ -- YuQing <384681@qq.com>  Wed, 31 Jan 2024 12:00:10 +0000
+
+libserverframe (1.2.1-1) unstable; urgency=medium
+
+  * upgrade to 1.2.1-1
+
+ -- YuQing <384681@qq.com>  Mon, 01 Jan 2024 11:24:45 +0000
+
+libserverframe (1.2.0-3) unstable; urgency=medium
+
+  * upgrade to 1.2.0-3
+
+ -- YuQing <384681@qq.com>  Tue, 21 Nov 2023 14:36:16 +0000
+
+libserverframe (1.2.0-2) unstable; urgency=medium
+
+  * upgrade to 1.2.0-2
+
+ -- YuQing <384681@qq.com>  Mon, 20 Nov 2023 13:24:02 +0000
+
+libserverframe (1.2.0-1) unstable; urgency=medium
+
+  * upgrade to 1.2.0-1
+
+ -- YuQing <384681@qq.com>  Sun, 19 Nov 2023 14:46:16 +0000
+
+libserverframe (1.1.29-1) unstable; urgency=medium
+
+  * upgrade to 1.1.29-1
+
+ -- YuQing <384681@qq.com>  Sun, 06 Aug 2023 07:22:46 +0000
+
+libserverframe (1.1.28-1) unstable; urgency=medium
+
+  * upgrade to 1.1.28-1
+
+ -- YuQing <384681@qq.com>  Sun, 23 Jul 2023 14:28:20 +0000
+
+libserverframe (1.1.27-1) unstable; urgency=medium
+
+  * upgrade to 1.1.27-1
+
+ -- YuQing <384681@qq.com>  Sat, 24 Jun 2023 06:51:30 +0000
+
+libserverframe (1.1.26-1) unstable; urgency=medium
+
+  * upgrade to 1.1.26-1
+
+ -- YuQing <384681@qq.com>  Sun, 04 Jun 2023 10:52:11 +0000
+
+libserverframe (1.1.25-1) unstable; urgency=medium
+
+  * upgrade to 1.1.25-1
+
+ -- YuQing <384681@qq.com>  Sat, 18 Feb 2023 05:44:50 +0000
+
+libserverframe (1.1.24-1) unstable; urgency=medium
+
+  * upgrade to 1.1.24-1
+
+ -- YuQing <384681@qq.com>  Sun, 15 Jan 2023 13:50:15 +0000
+
+libserverframe (1.1.22-1) unstable; urgency=medium
+
+  * upgrade to 1.1.22-1
+
+ -- YuQing <384681@qq.com>  Mon, 21 Nov 2022 14:55:50 +0000
+
+libserverframe (1.1.21-1) unstable; urgency=medium
+
+  * upgrade to 1.1.21-1
+
+ -- YuQing <384681@qq.com>  Sat, 08 Oct 2022 13:28:40 +0000
+
+libserverframe (1.1.20-1) unstable; urgency=medium
+
+  * upgrade to 1.1.20-1
+
+ -- YuQing <384681@qq.com>  Thu, 22 Sep 2022 12:22:39 +0000
+
+libserverframe (1.1.18-1) unstable; urgency=medium
+
+  * upgrade to 1.1.18-1
+
+ -- YuQing <384681@qq.com>  Wed, 07 Sep 2022 13:36:38 +0000
+
+libserverframe (1.1.17-1) unstable; urgency=medium
+
+  * upgrade to 1.1.17-1
+
+ -- YuQing <384681@qq.com>  Mon, 15 Aug 2022 13:31:54 +0000
+
+libserverframe (1.1.16-1) unstable; urgency=medium
+
+  * upgrade to 1.1.16-1
+
+ -- YuQing <384681@qq.com>  Mon, 25 Jul 2022 13:52:09 +0000
+
+libserverframe (1.1.15-1) unstable; urgency=medium
+
+  * upgrade to 1.1.15-1
+
+ -- YuQing <384681@qq.com>  Wed, 15 Jun 2022 14:26:27 +0000
+
+libserverframe (1.1.14-1) unstable; urgency=medium
+
+  * upgrade to 1.1.14-1
+
+ -- YuQing <384681@qq.com>  Thu, 28 Apr 2022 11:54:26 +0000
+
+libserverframe (1.1.13-1) unstable; urgency=medium
+
+  * add files: sf_iov.[hc]
+  * iova_slice error detect
+  * sf_iova_memset impl.
+  * sf_iova_memset_ex for iov and iovcnt
+  * sf_iova_memset_ex: add const modifier
+  * sf_iov.[hc] add function sf_iova_memcpy_ex
+  * simple_hash rename to fc_simple_hash
+  * change log level to debug
+  * add function sf_load_global_base_path
+  * support function sf_sharding_htable_delete
+  * sf_load_global_config_ex: server_name can be NULL
+  * upgrade version to 1.1.13
+  * make.sh: change DEBUG_FLAG to 0
+  * add function sf_binlog_writer_get_index_filename
+
+ -- YuQing <384681@qq.com>  Sun, 13 Mar 2022 16:46:17 +0800
+
+libserverframe (1.1.12-1) unstable; urgency=medium
+
+  * upgrade version to 1.1.12
+
+ -- YuQing <384681@qq.com>  Sat, 15 Jan 2022 20:00:21 +0800
+
+libserverframe (1.1.11-1) unstable; urgency=medium
+
+  * upgrade version to 1.1.11
+
+ --  YuQing <384681@qq.com>  Sun, 26 Dec 2021 21:02:05 +0800
+
+libserverframe (1.1.10-1) unstable; urgency=medium
+
+  * fixed somebugs
+
+ --  YuQing <384681@qq.com>  Sat, 10 Jul 2021 22:29:13 +0800

debian/compat

@@ -0,0 +1 @@
+11

debian/control

@@ -0,0 +1,25 @@
+Source: libserverframe
+Section: net
+Priority: optional
+Maintainer: YuQing <384681@qq.com>
+Build-Depends: debhelper (>=11~)
+ , libfastcommon-dev (>= 1.0.56)
+Standards-Version: 4.1.4
+Homepage: https://github.com/happyfish100/libserverframe
+
+Package: libserverframe-dev
+Section: net
+Architecture: any
+Multi-Arch: same
+Pre-Depends: ${misc:Pre-Depends}
+Depends: ${misc:Depends}, libserverframe (= ${binary:Version})
+Description: libserverframe (development files)
+  This package contains header files.
+
+Package: libserverframe
+Section: net
+Architecture: any
+Multi-Arch: same
+Pre-Depends: ${misc:Pre-Depends}
+Depends: ${misc:Depends}, ${shlibs:Depends}, libfastcommon (>= ${libfastcommon:Version})
+Description: this network service framework library extract from FastDFS

debian/copyright

@@ -0,0 +1,680 @@
+Format: https://www.debian.org/doc/packaging-manuals/copyright-format/1.0/
+Upstream-Name: libserverframe
+Source: https://github.com/happyfish100/libserverframe
+
+Files:     *
+Copyright: 2020 YuQing <384681@qq.com>
+License:   AGPL-3.0+
+ This program is free software: you can use, redistribute, and/or modify
+ it under the terms of the GNU Affero General Public License, version 3
+ or later ("AGPL"), as published by the Free Software Foundation.
+ .
+ This program is distributed in the hope that it will be useful, but WITHOUT
+ ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ FITNESS FOR A PARTICULAR PURPOSE.
+ .
+ You should have received a copy of the GNU Affero General Public License
+ along with this program. If not, see <https://www.gnu.org/licenses/>.
+
+# License file: LICENSE
+                     GNU AFFERO GENERAL PUBLIC LICENSE
+                        Version 3, 19 November 2007
+ .
+  Copyright (C) 2007 Free Software Foundation, Inc. <https://fsf.org/>
+  Everyone is permitted to copy and distribute verbatim copies
+  of this license document, but changing it is not allowed.
+ .
+                             Preamble
+ .
+   The GNU Affero General Public License is a free, copyleft license for
+ software and other kinds of works, specifically designed to ensure
+ cooperation with the community in the case of network server software.
+ .
+   The licenses for most software and other practical works are designed
+ to take away your freedom to share and change the works.  By contrast,
+ our General Public Licenses are intended to guarantee your freedom to
+ share and change all versions of a program--to make sure it remains free
+ software for all its users.
+ .
+   When we speak of free software, we are referring to freedom, not
+ price.  Our General Public Licenses are designed to make sure that you
+ have the freedom to distribute copies of free software (and charge for
+ them if you wish), that you receive source code or can get it if you
+ want it, that you can change the software or use pieces of it in new
+ free programs, and that you know you can do these things.
+ .
+   Developers that use our General Public Licenses protect your rights
+ with two steps: (1) assert copyright on the software, and (2) offer
+ you this License which gives you legal permission to copy, distribute
+ and/or modify the software.
+ .
+   A secondary benefit of defending all users' freedom is that
+ improvements made in alternate versions of the program, if they
+ receive widespread use, become available for other developers to
+ incorporate.  Many developers of free software are heartened and
+ encouraged by the resulting cooperation.  However, in the case of
+ software used on network servers, this result may fail to come about.
+ The GNU General Public License permits making a modified version and
+ letting the public access it on a server without ever releasing its
+ source code to the public.
+ .
+   The GNU Affero General Public License is designed specifically to
+ ensure that, in such cases, the modified source code becomes available
+ to the community.  It requires the operator of a network server to
+ provide the source code of the modified version running there to the
+ users of that server.  Therefore, public use of a modified version, on
+ a publicly accessible server, gives the public access to the source
+ code of the modified version.
+ .
+   An older license, called the Affero General Public License and
+ published by Affero, was designed to accomplish similar goals.  This is
+ a different license, not a version of the Affero GPL, but Affero has
+ released a new version of the Affero GPL which permits relicensing under
+ this license.
+ .
+   The precise terms and conditions for copying, distribution and
+ modification follow.
+ .
+                        TERMS AND CONDITIONS
+ .
+   0. Definitions.
+ .
+   "This License" refers to version 3 of the GNU Affero General Public License.
+ .
+   "Copyright" also means copyright-like laws that apply to other kinds of
+ works, such as semiconductor masks.
+ .
+   "The Program" refers to any copyrightable work licensed under this
+ License.  Each licensee is addressed as "you".  "Licensees" and
+ "recipients" may be individuals or organizations.
+ .
+   To "modify" a work means to copy from or adapt all or part of the work
+ in a fashion requiring copyright permission, other than the making of an
+ exact copy.  The resulting work is called a "modified version" of the
+ earlier work or a work "based on" the earlier work.
+ .
+   A "covered work" means either the unmodified Program or a work based
+ on the Program.
+ .
+   To "propagate" a work means to do anything with it that, without
+ permission, would make you directly or secondarily liable for
+ infringement under applicable copyright law, except executing it on a
+ computer or modifying a private copy.  Propagation includes copying,
+ distribution (with or without modification), making available to the
+ public, and in some countries other activities as well.
+ .
+   To "convey" a work means any kind of propagation that enables other
+ parties to make or receive copies.  Mere interaction with a user through
+ a computer network, with no transfer of a copy, is not conveying.
+ .
+   An interactive user interface displays "Appropriate Legal Notices"
+ to the extent that it includes a convenient and prominently visible
+ feature that (1) displays an appropriate copyright notice, and (2)
+ tells the user that there is no warranty for the work (except to the
+ extent that warranties are provided), that licensees may convey the
+ work under this License, and how to view a copy of this License.  If
+ the interface presents a list of user commands or options, such as a
+ menu, a prominent item in the list meets this criterion.
+ .
+   1. Source Code.
+ .
+   The "source code" for a work means the preferred form of the work
+ for making modifications to it.  "Object code" means any non-source
+ form of a work.
+ .
+   A "Standard Interface" means an interface that either is an official
+ standard defined by a recognized standards body, or, in the case of
+ interfaces specified for a particular programming language, one that
+ is widely used among developers working in that language.
+ .
+   The "System Libraries" of an executable work include anything, other
+ than the work as a whole, that (a) is included in the normal form of
+ packaging a Major Component, but which is not part of that Major
+ Component, and (b) serves only to enable use of the work with that
+ Major Component, or to implement a Standard Interface for which an
+ implementation is available to the public in source code form.  A
+ "Major Component", in this context, means a major essential component
+ (kernel, window system, and so on) of the specific operating system
+ (if any) on which the executable work runs, or a compiler used to
+ produce the work, or an object code interpreter used to run it.
+ .
+   The "Corresponding Source" for a work in object code form means all
+ the source code needed to generate, install, and (for an executable
+ work) run the object code and to modify the work, including scripts to
+ control those activities.  However, it does not include the work's
+ System Libraries, or general-purpose tools or generally available free
+ programs which are used unmodified in performing those activities but
+ which are not part of the work.  For example, Corresponding Source
+ includes interface definition files associated with source files for
+ the work, and the source code for shared libraries and dynamically
+ linked subprograms that the work is specifically designed to require,
+ such as by intimate data communication or control flow between those
+ subprograms and other parts of the work.
+ .
+   The Corresponding Source need not include anything that users
+ can regenerate automatically from other parts of the Corresponding
+ Source.
+ .
+   The Corresponding Source for a work in source code form is that
+ same work.
+ .
+   2. Basic Permissions.
+ .
+   All rights granted under this License are granted for the term of
+ copyright on the Program, and are irrevocable provided the stated
+ conditions are met.  This License explicitly affirms your unlimited
+ permission to run the unmodified Program.  The output from running a
+ covered work is covered by this License only if the output, given its
+ content, constitutes a covered work.  This License acknowledges your
+ rights of fair use or other equivalent, as provided by copyright law.
+ .
+   You may make, run and propagate covered works that you do not
+ convey, without conditions so long as your license otherwise remains
+ in force.  You may convey covered works to others for the sole purpose
+ of having them make modifications exclusively for you, or provide you
+ with facilities for running those works, provided that you comply with
+ the terms of this License in conveying all material for which you do
+ not control copyright.  Those thus making or running the covered works
+ for you must do so exclusively on your behalf, under your direction
+ and control, on terms that prohibit them from making any copies of
+ your copyrighted material outside their relationship with you.
+ .
+   Conveying under any other circumstances is permitted solely under
+ the conditions stated below.  Sublicensing is not allowed; section 10
+ makes it unnecessary.
+ .
+   3. Protecting Users' Legal Rights From Anti-Circumvention Law.
+ .
+   No covered work shall be deemed part of an effective technological
+ measure under any applicable law fulfilling obligations under article
+ 11 of the WIPO copyright treaty adopted on 20 December 1996, or
+ similar laws prohibiting or restricting circumvention of such
+ measures.
+ .
+   When you convey a covered work, you waive any legal power to forbid
+ circumvention of technological measures to the extent such circumvention
+ is effected by exercising rights under this License with respect to
+ the covered work, and you disclaim any intention to limit operation or
+ modification of the work as a means of enforcing, against the work's
+ users, your or third parties' legal rights to forbid circumvention of
+ technological measures.
+ .
+   4. Conveying Verbatim Copies.
+ .
+   You may convey verbatim copies of the Program's source code as you
+ receive it, in any medium, provided that you conspicuously and
+ appropriately publish on each copy an appropriate copyright notice;
+ keep intact all notices stating that this License and any
+ non-permissive terms added in accord with section 7 apply to the code;
+ keep intact all notices of the absence of any warranty; and give all
+ recipients a copy of this License along with the Program.
+ .
+   You may charge any price or no price for each copy that you convey,
+ and you may offer support or warranty protection for a fee.
+ .
+   5. Conveying Modified Source Versions.
+ .
+   You may convey a work based on the Program, or the modifications to
+ produce it from the Program, in the form of source code under the
+ terms of section 4, provided that you also meet all of these conditions:
+ .
+     a) The work must carry prominent notices stating that you modified
+     it, and giving a relevant date.
+ .
+     b) The work must carry prominent notices stating that it is
+     released under this License and any conditions added under section
+     7.  This requirement modifies the requirement in section 4 to
+     "keep intact all notices".
+ .
+     c) You must license the entire work, as a whole, under this
+     License to anyone who comes into possession of a copy.  This
+     License will therefore apply, along with any applicable section 7
+     additional terms, to the whole of the work, and all its parts,
+     regardless of how they are packaged.  This License gives no
+     permission to license the work in any other way, but it does not
+     invalidate such permission if you have separately received it.
+ .
+     d) If the work has interactive user interfaces, each must display
+     Appropriate Legal Notices; however, if the Program has interactive
+     interfaces that do not display Appropriate Legal Notices, your
+     work need not make them do so.
+ .
+   A compilation of a covered work with other separate and independent
+ works, which are not by their nature extensions of the covered work,
+ and which are not combined with it such as to form a larger program,
+ in or on a volume of a storage or distribution medium, is called an
+ "aggregate" if the compilation and its resulting copyright are not
+ used to limit the access or legal rights of the compilation's users
+ beyond what the individual works permit.  Inclusion of a covered work
+ in an aggregate does not cause this License to apply to the other
+ parts of the aggregate.
+ .
+   6. Conveying Non-Source Forms.
+ .
+   You may convey a covered work in object code form under the terms
+ of sections 4 and 5, provided that you also convey the
+ machine-readable Corresponding Source under the terms of this License,
+ in one of these ways:
+ .
+     a) Convey the object code in, or embodied in, a physical product
+     (including a physical distribution medium), accompanied by the
+     Corresponding Source fixed on a durable physical medium
+     customarily used for software interchange.
+ .
+     b) Convey the object code in, or embodied in, a physical product
+     (including a physical distribution medium), accompanied by a
+     written offer, valid for at least three years and valid for as
+     long as you offer spare parts or customer support for that product
+     model, to give anyone who possesses the object code either (1) a
+     copy of the Corresponding Source for all the software in the
+     product that is covered by this License, on a durable physical
+     medium customarily used for software interchange, for a price no
+     more than your reasonable cost of physically performing this
+     conveying of source, or (2) access to copy the
+     Corresponding Source from a network server at no charge.
+ .
+     c) Convey individual copies of the object code with a copy of the
+     written offer to provide the Corresponding Source.  This
+     alternative is allowed only occasionally and noncommercially, and
+     only if you received the object code with such an offer, in accord
+     with subsection 6b.
+ .
+     d) Convey the object code by offering access from a designated
+     place (gratis or for a charge), and offer equivalent access to the
+     Corresponding Source in the same way through the same place at no
+     further charge.  You need not require recipients to copy the
+     Corresponding Source along with the object code.  If the place to
+     copy the object code is a network server, the Corresponding Source
+     may be on a different server (operated by you or a third party)
+     that supports equivalent copying facilities, provided you maintain
+     clear directions next to the object code saying where to find the
+     Corresponding Source.  Regardless of what server hosts the
+     Corresponding Source, you remain obligated to ensure that it is
+     available for as long as needed to satisfy these requirements.
+ .
+     e) Convey the object code using peer-to-peer transmission, provided
+     you inform other peers where the object code and Corresponding
+     Source of the work are being offered to the general public at no
+     charge under subsection 6d.
+ .
+   A separable portion of the object code, whose source code is excluded
+ from the Corresponding Source as a System Library, need not be
+ included in conveying the object code work.
+ .
+   A "User Product" is either (1) a "consumer product", which means any
+ tangible personal property which is normally used for personal, family,
+ or household purposes, or (2) anything designed or sold for incorporation
+ into a dwelling.  In determining whether a product is a consumer product,
+ doubtful cases shall be resolved in favor of coverage.  For a particular
+ product received by a particular user, "normally used" refers to a
+ typical or common use of that class of product, regardless of the status
+ of the particular user or of the way in which the particular user
+ actually uses, or expects or is expected to use, the product.  A product
+ is a consumer product regardless of whether the product has substantial
+ commercial, industrial or non-consumer uses, unless such uses represent
+ the only significant mode of use of the product.
+ .
+   "Installation Information" for a User Product means any methods,
+ procedures, authorization keys, or other information required to install
+ and execute modified versions of a covered work in that User Product from
+ a modified version of its Corresponding Source.  The information must
+ suffice to ensure that the continued functioning of the modified object
+ code is in no case prevented or interfered with solely because
+ modification has been made.
+ .
+   If you convey an object code work under this section in, or with, or
+ specifically for use in, a User Product, and the conveying occurs as
+ part of a transaction in which the right of possession and use of the
+ User Product is transferred to the recipient in perpetuity or for a
+ fixed term (regardless of how the transaction is characterized), the
+ Corresponding Source conveyed under this section must be accompanied
+ by the Installation Information.  But this requirement does not apply
+ if neither you nor any third party retains the ability to install
+ modified object code on the User Product (for example, the work has
+ been installed in ROM).
+ .
+   The requirement to provide Installation Information does not include a
+ requirement to continue to provide support service, warranty, or updates
+ for a work that has been modified or installed by the recipient, or for
+ the User Product in which it has been modified or installed.  Access to a
+ network may be denied when the modification itself materially and
+ adversely affects the operation of the network or violates the rules and
+ protocols for communication across the network.
+ .
+   Corresponding Source conveyed, and Installation Information provided,
+ in accord with this section must be in a format that is publicly
+ documented (and with an implementation available to the public in
+ source code form), and must require no special password or key for
+ unpacking, reading or copying.
+ .
+   7. Additional Terms.
+ .
+   "Additional permissions" are terms that supplement the terms of this
+ License by making exceptions from one or more of its conditions.
+ Additional permissions that are applicable to the entire Program shall
+ be treated as though they were included in this License, to the extent
+ that they are valid under applicable law.  If additional permissions
+ apply only to part of the Program, that part may be used separately
+ under those permissions, but the entire Program remains governed by
+ this License without regard to the additional permissions.
+ .
+   When you convey a copy of a covered work, you may at your option
+ remove any additional permissions from that copy, or from any part of
+ it.  (Additional permissions may be written to require their own
+ removal in certain cases when you modify the work.)  You may place
+ additional permissions on material, added by you to a covered work,
+ for which you have or can give appropriate copyright permission.
+ .
+   Notwithstanding any other provision of this License, for material you
+ add to a covered work, you may (if authorized by the copyright holders of
+ that material) supplement the terms of this License with terms:
+ .
+     a) Disclaiming warranty or limiting liability differently from the
+     terms of sections 15 and 16 of this License; or
+ .
+     b) Requiring preservation of specified reasonable legal notices or
+     author attributions in that material or in the Appropriate Legal
+     Notices displayed by works containing it; or
+ .
+     c) Prohibiting misrepresentation of the origin of that material, or
+     requiring that modified versions of such material be marked in
+     reasonable ways as different from the original version; or
+ .
+     d) Limiting the use for publicity purposes of names of licensors or
+     authors of the material; or
+ .
+     e) Declining to grant rights under trademark law for use of some
+     trade names, trademarks, or service marks; or
+ .
+     f) Requiring indemnification of licensors and authors of that
+     material by anyone who conveys the material (or modified versions of
+     it) with contractual assumptions of liability to the recipient, for
+     any liability that these contractual assumptions directly impose on
+     those licensors and authors.
+ .
+   All other non-permissive additional terms are considered "further
+ restrictions" within the meaning of section 10.  If the Program as you
+ received it, or any part of it, contains a notice stating that it is
+ governed by this License along with a term that is a further
+ restriction, you may remove that term.  If a license document contains
+ a further restriction but permits relicensing or conveying under this
+ License, you may add to a covered work material governed by the terms
+ of that license document, provided that the further restriction does
+ not survive such relicensing or conveying.
+ .
+   If you add terms to a covered work in accord with this section, you
+ must place, in the relevant source files, a statement of the
+ additional terms that apply to those files, or a notice indicating
+ where to find the applicable terms.
+ .
+   Additional terms, permissive or non-permissive, may be stated in the
+ form of a separately written license, or stated as exceptions;
+ the above requirements apply either way.
+ .
+   8. Termination.
+ .
+   You may not propagate or modify a covered work except as expressly
+ provided under this License.  Any attempt otherwise to propagate or
+ modify it is void, and will automatically terminate your rights under
+ this License (including any patent licenses granted under the third
+ paragraph of section 11).
+ .
+   However, if you cease all violation of this License, then your
+ license from a particular copyright holder is reinstated (a)
+ provisionally, unless and until the copyright holder explicitly and
+ finally terminates your license, and (b) permanently, if the copyright
+ holder fails to notify you of the violation by some reasonable means
+ prior to 60 days after the cessation.
+ .
+   Moreover, your license from a particular copyright holder is
+ reinstated permanently if the copyright holder notifies you of the
+ violation by some reasonable means, this is the first time you have
+ received notice of violation of this License (for any work) from that
+ copyright holder, and you cure the violation prior to 30 days after
+ your receipt of the notice.
+ .
+   Termination of your rights under this section does not terminate the
+ licenses of parties who have received copies or rights from you under
+ this License.  If your rights have been terminated and not permanently
+ reinstated, you do not qualify to receive new licenses for the same
+ material under section 10.
+ .
+   9. Acceptance Not Required for Having Copies.
+ .
+   You are not required to accept this License in order to receive or
+ run a copy of the Program.  Ancillary propagation of a covered work
+ occurring solely as a consequence of using peer-to-peer transmission
+ to receive a copy likewise does not require acceptance.  However,
+ nothing other than this License grants you permission to propagate or
+ modify any covered work.  These actions infringe copyright if you do
+ not accept this License.  Therefore, by modifying or propagating a
+ covered work, you indicate your acceptance of this License to do so.
+ .
+   10. Automatic Licensing of Downstream Recipients.
+ .
+   Each time you convey a covered work, the recipient automatically
+ receives a license from the original licensors, to run, modify and
+ propagate that work, subject to this License.  You are not responsible
+ for enforcing compliance by third parties with this License.
+ .
+   An "entity transaction" is a transaction transferring control of an
+ organization, or substantially all assets of one, or subdividing an
+ organization, or merging organizations.  If propagation of a covered
+ work results from an entity transaction, each party to that
+ transaction who receives a copy of the work also receives whatever
+ licenses to the work the party's predecessor in interest had or could
+ give under the previous paragraph, plus a right to possession of the
+ Corresponding Source of the work from the predecessor in interest, if
+ the predecessor has it or can get it with reasonable efforts.
+ .
+   You may not impose any further restrictions on the exercise of the
+ rights granted or affirmed under this License.  For example, you may
+ not impose a license fee, royalty, or other charge for exercise of
+ rights granted under this License, and you may not initiate litigation
+ (including a cross-claim or counterclaim in a lawsuit) alleging that
+ any patent claim is infringed by making, using, selling, offering for
+ sale, or importing the Program or any portion of it.
+ .
+   11. Patents.
+ .
+   A "contributor" is a copyright holder who authorizes use under this
+ License of the Program or a work on which the Program is based.  The
+ work thus licensed is called the contributor's "contributor version".
+ .
+   A contributor's "essential patent claims" are all patent claims
+ owned or controlled by the contributor, whether already acquired or
+ hereafter acquired, that would be infringed by some manner, permitted
+ by this License, of making, using, or selling its contributor version,
+ but do not include claims that would be infringed only as a
+ consequence of further modification of the contributor version.  For
+ purposes of this definition, "control" includes the right to grant
+ patent sublicenses in a manner consistent with the requirements of
+ this License.
+ .
+   Each contributor grants you a non-exclusive, worldwide, royalty-free
+ patent license under the contributor's essential patent claims, to
+ make, use, sell, offer for sale, import and otherwise run, modify and
+ propagate the contents of its contributor version.
+ .
+   In the following three paragraphs, a "patent license" is any express
+ agreement or commitment, however denominated, not to enforce a patent
+ (such as an express permission to practice a patent or covenant not to
+ sue for patent infringement).  To "grant" such a patent license to a
+ party means to make such an agreement or commitment not to enforce a
+ patent against the party.
+ .
+   If you convey a covered work, knowingly relying on a patent license,
+ and the Corresponding Source of the work is not available for anyone
+ to copy, free of charge and under the terms of this License, through a
+ publicly available network server or other readily accessible means,
+ then you must either (1) cause the Corresponding Source to be so
+ available, or (2) arrange to deprive yourself of the benefit of the
+ patent license for this particular work, or (3) arrange, in a manner
+ consistent with the requirements of this License, to extend the patent
+ license to downstream recipients.  "Knowingly relying" means you have
+ actual knowledge that, but for the patent license, your conveying the
+ covered work in a country, or your recipient's use of the covered work
+ in a country, would infringe one or more identifiable patents in that
+ country that you have reason to believe are valid.
+ .
+   If, pursuant to or in connection with a single transaction or
+ arrangement, you convey, or propagate by procuring conveyance of, a
+ covered work, and grant a patent license to some of the parties
+ receiving the covered work authorizing them to use, propagate, modify
+ or convey a specific copy of the covered work, then the patent license
+ you grant is automatically extended to all recipients of the covered
+ work and works based on it.
+ .
+   A patent license is "discriminatory" if it does not include within
+ the scope of its coverage, prohibits the exercise of, or is
+ conditioned on the non-exercise of one or more of the rights that are
+ specifically granted under this License.  You may not convey a covered
+ work if you are a party to an arrangement with a third party that is
+ in the business of distributing software, under which you make payment
+ to the third party based on the extent of your activity of conveying
+ the work, and under which the third party grants, to any of the
+ parties who would receive the covered work from you, a discriminatory
+ patent license (a) in connection with copies of the covered work
+ conveyed by you (or copies made from those copies), or (b) primarily
+ for and in connection with specific products or compilations that
+ contain the covered work, unless you entered into that arrangement,
+ or that patent license was granted, prior to 28 March 2007.
+ .
+   Nothing in this License shall be construed as excluding or limiting
+ any implied license or other defenses to infringement that may
+ otherwise be available to you under applicable patent law.
+ .
+   12. No Surrender of Others' Freedom.
+ .
+   If conditions are imposed on you (whether by court order, agreement or
+ otherwise) that contradict the conditions of this License, they do not
+ excuse you from the conditions of this License.  If you cannot convey a
+ covered work so as to satisfy simultaneously your obligations under this
+ License and any other pertinent obligations, then as a consequence you may
+ not convey it at all.  For example, if you agree to terms that obligate you
+ to collect a royalty for further conveying from those to whom you convey
+ the Program, the only way you could satisfy both those terms and this
+ License would be to refrain entirely from conveying the Program.
+ .
+   13. Remote Network Interaction; Use with the GNU General Public License.
+ .
+   Notwithstanding any other provision of this License, if you modify the
+ Program, your modified version must prominently offer all users
+ interacting with it remotely through a computer network (if your version
+ supports such interaction) an opportunity to receive the Corresponding
+ Source of your version by providing access to the Corresponding Source
+ from a network server at no charge, through some standard or customary
+ means of facilitating copying of software.  This Corresponding Source
+ shall include the Corresponding Source for any work covered by version 3
+ of the GNU General Public License that is incorporated pursuant to the
+ following paragraph.
+ .
+   Notwithstanding any other provision of this License, you have
+ permission to link or combine any covered work with a work licensed
+ under version 3 of the GNU General Public License into a single
+ combined work, and to convey the resulting work.  The terms of this
+ License will continue to apply to the part which is the covered work,
+ but the work with which it is combined will remain governed by version
+ 3 of the GNU General Public License.
+ .
+   14. Revised Versions of this License.
+ .
+   The Free Software Foundation may publish revised and/or new versions of
+ the GNU Affero General Public License from time to time.  Such new versions
+ will be similar in spirit to the present version, but may differ in detail to
+ address new problems or concerns.
+ .
+   Each version is given a distinguishing version number.  If the
+ Program specifies that a certain numbered version of the GNU Affero General
+ Public License "or any later version" applies to it, you have the
+ option of following the terms and conditions either of that numbered
+ version or of any later version published by the Free Software
+ Foundation.  If the Program does not specify a version number of the
+ GNU Affero General Public License, you may choose any version ever published
+ by the Free Software Foundation.
+ .
+   If the Program specifies that a proxy can decide which future
+ versions of the GNU Affero General Public License can be used, that proxy's
+ public statement of acceptance of a version permanently authorizes you
+ to choose that version for the Program.
+ .
+   Later license versions may give you additional or different
+ permissions.  However, no additional obligations are imposed on any
+ author or copyright holder as a result of your choosing to follow a
+ later version.
+ .
+   15. Disclaimer of Warranty.
+ .
+   THERE IS NO WARRANTY FOR THE PROGRAM, TO THE EXTENT PERMITTED BY
+ APPLICABLE LAW.  EXCEPT WHEN OTHERWISE STATED IN WRITING THE COPYRIGHT
+ HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM "AS IS" WITHOUT WARRANTY
+ OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO,
+ THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+ PURPOSE.  THE ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAM
+ IS WITH YOU.  SHOULD THE PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF
+ ALL NECESSARY SERVICING, REPAIR OR CORRECTION.
+ .
+   16. Limitation of Liability.
+ .
+   IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING
+ WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MODIFIES AND/OR CONVEYS
+ THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES, INCLUDING ANY
+ GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING OUT OF THE
+ USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED TO LOSS OF
+ DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD
+ PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER PROGRAMS),
+ EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE POSSIBILITY OF
+ SUCH DAMAGES.
+ .
+   17. Interpretation of Sections 15 and 16.
+ .
+   If the disclaimer of warranty and limitation of liability provided
+ above cannot be given local legal effect according to their terms,
+ reviewing courts shall apply local law that most closely approximates
+ an absolute waiver of all civil liability in connection with the
+ Program, unless a warranty or assumption of liability accompanies a
+ copy of the Program in return for a fee.
+ .
+                      END OF TERMS AND CONDITIONS
+ .
+             How to Apply These Terms to Your New Programs
+ .
+   If you develop a new program, and you want it to be of the greatest
+ possible use to the public, the best way to achieve this is to make it
+ free software which everyone can redistribute and change under these terms.
+ .
+   To do so, attach the following notices to the program.  It is safest
+ to attach them to the start of each source file to most effectively
+ state the exclusion of warranty; and each file should have at least
+ the "copyright" line and a pointer to where the full notice is found.
+ .
+     <one line to give the program's name and a brief idea of what it does.>
+     Copyright (C) <year>  <name of author>
+ .
+     This program is free software: you can redistribute it and/or modify
+     it under the terms of the GNU Affero General Public License as published
+     by the Free Software Foundation, either version 3 of the License, or
+     (at your option) any later version.
+ .
+     This program is distributed in the hope that it will be useful,
+     but WITHOUT ANY WARRANTY; without even the implied warranty of
+     MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+     GNU Affero General Public License for more details.
+ .
+     You should have received a copy of the GNU Affero General Public License
+     along with this program.  If not, see <https://www.gnu.org/licenses/>.
+ .
+ Also add information on how to contact you by electronic and paper mail.
+ .
+   If your software can interact with users remotely through a computer
+ network, you should also make sure that it provides a way for users to
+ get its source.  For example, if your program is a web application, its
+ interface could display a "Source" link that leads users to an archive
+ of the code.  There are many ways you could offer source, and different
+ solutions will be better for different programs; see section 13 for the
+ specific requirements.
+ .
+   You should also get your employer (if you work as a programmer) or school,
+ if any, to sign a "copyright disclaimer" for the program, if necessary.
+ For more information on this, and how to apply and follow the GNU AGPL, see
+ <https://www.gnu.org/licenses/>.

debian/libserverframe-dev.install

@@ -0,0 +1 @@
+usr/include/sf/*

debian/libserverframe.install

@@ -0,0 +1 @@
+usr/lib/libserverframe.so*

debian/rules

@@ -0,0 +1,17 @@
+#!/usr/bin/make -f
+
+export DESTDIR = $(CURDIR)/debian/tmp
+
+%:
+	dh $@
+
+override_dh_auto_build:
+	./make.sh clean && ./make.sh
+
+override_dh_auto_install:
+	./make.sh install
+	dh_auto_install
+
+.PHONY: override_dh_gencontrol
+override_dh_gencontrol:
+	dh_gencontrol -- -Tdebian/substvars

debian/source/format

@@ -0,0 +1 @@
+3.0 (quilt)

debian/substvars

@@ -0,0 +1 @@
+libfastcommon:Version=1.0.83

debian/watch

@@ -0,0 +1,3 @@
+version=3
+opts="mode=git" https://github.com/happyfish100/libserverframe.git \
+   refs/tags/v([\d\.]+) debian uupdate

libserverframe.spec

@@ -2,19 +2,20 @@
 %define CommitVersion %(echo $COMMIT_VERSION)
 
 Name: libserverframe
-Version: 1.1.0
+Version: 1.2.11
 Release: 1%{?dist}
-Summary: mc common framework library
-License: GPL
+Summary: network framework library
+License: AGPL v3.0
 Group: Arch/Tech
-URL:  http://github.com/happyfish100/libfastcommon/
-Source: http://github.com/happyfish100/libfastcommon/%{name}-%{version}.tar.gz
+URL:  http://github.com/happyfish100/libserverframe/
+Source: http://github.com/happyfish100/libserverframe/%{name}-%{version}.tar.gz
 
-BuildRoot: %{_tmppath}/%{name}-%{version}-%{release}-root-%(%{__id_u} -n) 
+BuildRoot: %{_tmppath}/%{name}-%{version}-%{release}-root-%(%{__id_u} -n)
 
-BuildRequires: libfastcommon-devel >= 1.0.44
+BuildRequires: libfastcommon-devel >= 1.0.83
 Requires: %__cp %__mv %__chmod %__grep %__mkdir %__install %__id
-Requires: libfastcommon >= 1.0.44
+Requires: libfastcommon >= 1.0.83
+
 %description
 common framework library 
 commit version: %{CommitVersion}
@@ -32,7 +33,7 @@ commit version: %{CommitVersion}
 %setup -q
 
 %build
-./make.sh
+./make.sh clean && ./make.sh
 
 %install
 rm -rf %{buildroot}
@@ -56,5 +57,5 @@ rm -rf %{buildroot}
 /usr/include/sf/*
 
 %changelog
-* Mon Jun 23 2014  Zaixue Liao <liaozaixue@yongche.com>
+* Mon Jun 23 2014  Zaixue Liao
 - first RPM release (1.0)

make.sh

@@ -5,18 +5,32 @@ TARGET_PREFIX=$DESTDIR/usr
 TARGET_CONF_PATH=$DESTDIR/etc
 
 LIB_VERSION=lib64
-DEBUG_FLAG=1
+DEBUG_FLAG=0
 
 if [ -f /usr/include/fastcommon/_os_define.h ]; then
-  OS_BITS=$(fgrep OS_BITS /usr/include/fastcommon/_os_define.h | awk '{print $NF;}')
+  OS_BITS=$(grep -F OS_BITS /usr/include/fastcommon/_os_define.h | awk '{print $NF;}')
+  USE_URING=$(grep -F IOEVENT_USE_URING /usr/include/fastcommon/_os_define.h | awk '{print $NF;}')
 elif [ -f /usr/local/include/fastcommon/_os_define.h ]; then
-  OS_BITS=$(fgrep OS_BITS /usr/local/include/fastcommon/_os_define.h | awk '{print $NF;}')
+  OS_BITS=$(grep -F OS_BITS /usr/local/include/fastcommon/_os_define.h | awk '{print $NF;}')
+  USE_URING=$(grep -F IOEVENT_USE_URING /usr/local/include/fastcommon/_os_define.h | awk '{print $NF;}')
 else
   OS_BITS=64
+  USE_URING=''
 fi
 
+uname=$(uname)
 if [ "$OS_BITS" -eq 64 ]; then
-  LIB_VERSION=lib64
+  if [ $uname = 'Linux' ]; then
+    osname=$(cat /etc/os-release | grep -w NAME | awk -F '=' '{print $2;}' | \
+            awk -F '"' '{if (NF==3) {print $2} else {print $1}}' | awk '{print $1}')
+    if [ $osname = 'Ubuntu' -o $osname = 'Debian' ]; then
+      LIB_VERSION=lib
+    else
+      LIB_VERSION=lib64
+    fi
+  else
+    LIB_VERSION=lib
+  fi
 else
   LIB_VERSION=lib
 fi
@@ -31,13 +45,16 @@ CFLAGS="$CFLAGS -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE"
 if [ "$DEBUG_FLAG" = "1" ]; then
   CFLAGS="$CFLAGS -g -O1 -DDEBUG_FLAG"
 else
-  CFLAGS="$CFLAGS -O3"
+  CFLAGS="$CFLAGS -g -O3"
 fi
 
 LIBS=''
 uname=$(uname)
 if [ "$uname" = "Linux" ]; then
   CFLAGS="$CFLAGS"
+  if [ -n "$USE_URING" ]; then
+    LIBS="$LIBS -luring"
+  fi
 elif [ "$uname" = "FreeBSD" ] || [ "$uname" = "Darwin" ]; then
   CFLAGS="$CFLAGS"
   if [ "$uname" = "Darwin" ]; then
@@ -66,19 +83,19 @@ elif [ "$uname" = "HP-UX" ]; then
   fi
 elif [ "$uname" = "FreeBSD" ]; then
   if [ -f /usr/lib/libc_r.so ]; then
-    line=$(nm -D /usr/lib/libc_r.so | grep pthread_create | grep -w T)
+    line=$(nm -D /usr/lib/libc_r.so | grep -F pthread_create | grep -w T)
     if [ $? -eq 0 ]; then
       LIBS="$LIBS -lc_r"
       have_pthread=1
     fi
   elif [ -f /lib64/libc_r.so ]; then
-    line=$(nm -D /lib64/libc_r.so | grep pthread_create | grep -w T)
+    line=$(nm -D /lib64/libc_r.so | grep -F pthread_create | grep -w T)
     if [ $? -eq 0 ]; then
       LIBS="$LIBS -lc_r"
       have_pthread=1
     fi
   elif [ -f /usr/lib64/libc_r.so ]; then
-    line=$(nm -D /usr/lib64/libc_r.so | grep pthread_create | grep -w T)
+    line=$(nm -D /usr/lib64/libc_r.so | grep -F pthread_create | grep -w T)
     if [ $? -eq 0 ]; then
       LIBS="$LIBS -lc_r"
       have_pthread=1
@@ -87,7 +104,7 @@ elif [ "$uname" = "FreeBSD" ]; then
 fi
 
 if [ $have_pthread -eq 0 ] && [ "$uname" = "Linux" ]; then
-   /sbin/ldconfig -p | fgrep libpthread.so > /dev/null
+   /sbin/ldconfig -p | grep -w libpthread.so > /dev/null
    if [ $? -eq 0 ]; then
       LIBS="$LIBS -lpthread"
    else
@@ -113,8 +130,15 @@ sed_replace()
     fi
 }
 
-cd src
+cd src/include
+link=$(readlink sf)
+if [ $? -ne 0 ] || [ "$link" != '..' -a "$link" != '../' ]; then
+  ln -sf .. sf
+fi
+
+cd ..
 cp Makefile.in Makefile
+sed_replace "s#\\\$(CC)#gcc#g" Makefile
 sed_replace "s#\\\$(CFLAGS)#$CFLAGS#g" Makefile
 sed_replace "s#\\\$(LIBS)#$LIBS#g" Makefile
 sed_replace "s#\\\$(TARGET_PREFIX)#$TARGET_PREFIX#g" Makefile

src/Makefile.in

@@ -6,14 +6,20 @@ LIB_PATH = $(LIBS) -lfastcommon
 TARGET_LIB = $(TARGET_PREFIX)/$(LIB_VERSION)
 
 TOP_HEADERS = sf_types.h sf_global.h sf_define.h sf_nio.h sf_service.h \
-              sf_func.h sf_util.h sf_configs.h sf_proto.h sf_binlog_writer.h \
-              sf_sharding_htable.h
+              sf_func.h sf_util.h sf_configs.h sf_proto.h sf_cluster_cfg.h \
+              sf_sharding_htable.h sf_connection_manager.h sf_serializer.h \
+              sf_binlog_index.h sf_file_writer.h sf_binlog_writer.h \
+              sf_ordered_writer.h sf_buffered_writer.h sf_iov.h \
+              sf_shared_mbuffer.h
+
+IDEMP_COMMON_HEADER = idempotency/common/idempotency_types.h
 
 IDEMP_SERVER_HEADER = idempotency/server/server_types.h \
                       idempotency/server/server_channel.h  \
                       idempotency/server/request_htable.h  \
                       idempotency/server/channel_htable.h  \
-                      idempotency/server/server_handler.h
+                      idempotency/server/server_handler.h  \
+                      idempotency/server/request_metadata.h
 
 IDEMP_CLIENT_HEADER = idempotency/client/client_types.h    \
                       idempotency/client/receipt_handler.h \
@@ -22,14 +28,19 @@ IDEMP_CLIENT_HEADER = idempotency/client/client_types.h    \
 
 ALL_HEADERS  = $(TOP_HEADERS) $(IDEMP_SERVER_HEADER) $(IDEMP_CLIENT_HEADER)
 
-SHARED_OBJS = sf_nio.lo sf_service.lo sf_global.lo \
+SHARED_OBJS = sf_nio.lo sf_iov.lo sf_service.lo sf_global.lo \
         sf_func.lo sf_util.lo sf_configs.lo sf_proto.lo \
-        sf_binlog_writer.lo sf_sharding_htable.lo  \
-        idempotency/server/server_channel.lo  \
-        idempotency/server/request_htable.lo  \
-        idempotency/server/channel_htable.lo  \
-        idempotency/server/server_handler.lo  \
-        idempotency/client/receipt_handler.lo \
+        sf_sharding_htable.lo sf_cluster_cfg.lo   \
+        sf_connection_manager.lo sf_serializer.lo \
+        sf_binlog_index.lo sf_file_writer.lo  \
+        sf_binlog_writer.lo sf_ordered_writer.lo \
+        sf_shared_mbuffer.lo \
+        idempotency/server/server_channel.lo   \
+        idempotency/server/request_htable.lo   \
+        idempotency/server/channel_htable.lo   \
+        idempotency/server/server_handler.lo   \
+        idempotency/server/request_metadata.lo \
+        idempotency/client/receipt_handler.lo  \
         idempotency/client/client_channel.lo
 
 ALL_OBJS = $(SHARED_OBJS)
@@ -50,15 +61,18 @@ libserverframe.so: $(SHARED_OBJS)
 install:
 	mkdir -p $(TARGET_LIB)
 	mkdir -p $(TARGET_PREFIX)/lib
+	mkdir -p $(TARGET_PREFIX)/include/sf/idempotency/common
 	mkdir -p $(TARGET_PREFIX)/include/sf/idempotency/server
 	mkdir -p $(TARGET_PREFIX)/include/sf/idempotency/client
 
 	install -m 755 $(ALL_LIBS) $(TARGET_LIB)
 	cp -f $(TOP_HEADERS) $(TARGET_PREFIX)/include/sf
+	cp -f $(IDEMP_COMMON_HEADER) $(TARGET_PREFIX)/include/sf/idempotency/common
 	cp -f $(IDEMP_SERVER_HEADER) $(TARGET_PREFIX)/include/sf/idempotency/server
 	cp -f $(IDEMP_CLIENT_HEADER) $(TARGET_PREFIX)/include/sf/idempotency/client
 
-	if [ ! -e $(TARGET_PREFIX)/lib/libserverframe.so ]; then ln -s $(TARGET_LIB)/libserverframe.so $(TARGET_PREFIX)/lib/libserverframe.so; fi
+	@BUILDROOT=$$(echo "$(TARGET_PREFIX)" | grep BUILDROOT); \
+	if [ -z "$$BUILDROOT" ] && [ "$(TARGET_LIB)" != "$(TARGET_PREFIX)/lib" ]; then ln -sf $(TARGET_LIB)/libserverframe.so $(TARGET_PREFIX)/lib/libserverframe.so; fi
 clean:
 	rm -f $(ALL_OBJS) $(ALL_LIBS) $(ALL_PRGS)
 

src/idempotency/client/client_channel.c

@@ -31,11 +31,11 @@
 #include "fastcommon/pthread_func.h"
 #include "fastcommon/sched_thread.h"
 #include "fastcommon/fc_queue.h"
-#include "../../sf_util.h"
-#include "../../sf_func.h"
-#include "../../sf_nio.h"
-#include "../../sf_global.h"
-#include "../../sf_service.h"
+#include "sf/sf_util.h"
+#include "sf/sf_func.h"
+#include "sf/sf_nio.h"
+#include "sf/sf_global.h"
+#include "sf/sf_service.h"
 #include "client_channel.h"
 
 typedef struct {
@@ -129,13 +129,13 @@ static int idempotency_channel_alloc_init(void *element, void *args)
 
     channel = (IdempotencyClientChannel *)element;
     if ((result=fast_mblock_init_ex1(&channel->receipt_allocator,
-                    "idempotency_receipt", sizeof(IdempotencyClientReceipt),
+                    "idempotency-receipt", sizeof(IdempotencyClientReceipt),
                     1024, 0, NULL, NULL, true)) != 0)
     {
         return result;
     }
 
-    if ((result=init_pthread_lock_cond_pair(&channel->lc_pair)) != 0) {
+    if ((result=init_pthread_lock_cond_pair(&channel->lcp)) != 0) {
         return result;
     }
 
@@ -153,7 +153,7 @@ int client_channel_init(IniFullContext *ini_ctx)
     }
 
     if ((result=fast_mblock_init_ex1(&channel_context.channel_allocator,
-                    "channel_info", sizeof(IdempotencyClientChannel),
+                    "channel-info", sizeof(IdempotencyClientChannel),
                     64, 0, idempotency_channel_alloc_init, NULL, true)) != 0)
     {
         return result;
@@ -170,18 +170,31 @@ void client_channel_destroy()
 {
 }
 
-struct fast_task_info *alloc_channel_task(IdempotencyClientChannel *channel,
-        const uint32_t hash_code, const char *server_ip, const uint16_t port,
-        int *err_no)
+static struct fast_task_info *alloc_channel_task(IdempotencyClientChannel
+        *channel, const uint32_t hash_code, const FCCommunicationType comm_type,
+        const char *server_ip, const uint16_t port, int *err_no)
 {
     struct fast_task_info *task;
+    SFAddressFamilyHandler *fh;
+    SFNetworkHandler *handler;
 
-    if ((task=sf_alloc_init_task(&g_sf_context, -1)) == NULL) {
+    if (is_ipv6_addr(server_ip)) {
+        fh = g_sf_context.handlers + SF_IPV6_ADDRESS_FAMILY_INDEX;
+    } else {
+        fh = g_sf_context.handlers + SF_IPV4_ADDRESS_FAMILY_INDEX;
+    }
+
+    if (comm_type == fc_comm_type_sock) {
+        handler = fh->handlers + SF_SOCKET_NETWORK_HANDLER_INDEX;
+    } else {
+        handler = fh->handlers + SF_RDMACM_NETWORK_HANDLER_INDEX;
+    }
+    if ((task=sf_alloc_init_task(handler, -1)) == NULL) {
         *err_no = ENOMEM;
         return NULL;
     }
 
-    snprintf(task->server_ip, sizeof(task->server_ip), "%s", server_ip);
+    fc_safe_strcpy(task->server_ip, server_ip);
     task->port = port;
     task->arg = channel;
     task->thread_data = g_sf_context.thread_data +
@@ -190,7 +203,8 @@ struct fast_task_info *alloc_channel_task(IdempotencyClientChannel *channel,
     channel->last_connect_time = g_current_time;
     if ((*err_no=sf_nio_notify(task, SF_NIO_STAGE_CONNECT)) != 0) {
         channel->in_ioevent = 0;   //rollback
-        sf_release_task(task);
+        __sync_sub_and_fetch(&task->reffer_count, 1);
+        free_queue_push(task);
         return NULL;
     }
     return task;
@@ -200,6 +214,15 @@ int idempotency_client_channel_check_reconnect(
         IdempotencyClientChannel *channel)
 {
     int result;
+    char formatted_ip[FORMATTED_IP_SIZE];
+
+#if IOEVENT_USE_URING
+    struct fast_task_info *task;
+    task = channel->task;
+    if (SF_CTX->use_io_uring && FC_ATOMIC_GET(task->reffer_count) > 1) {
+        return 0;
+    }
+#endif
 
     if (!__sync_bool_compare_and_swap(&channel->in_ioevent, 0, 1)) {
         return 0;
@@ -210,11 +233,16 @@ int idempotency_client_channel_check_reconnect(
         channel->last_connect_time = g_current_time;
     }
 
-    logDebug("file: "__FILE__", line: %d, "
-            "trigger connect to server %s:%u",
-            __LINE__, channel->task->server_ip,
-            channel->task->port);
+    if (FC_LOG_BY_LEVEL(LOG_DEBUG)) {
+        format_ip_address(channel->task->server_ip, formatted_ip);
+        logDebug("file: "__FILE__", line: %d, "
+                "trigger connect to server %s:%u", __LINE__,
+                formatted_ip, channel->task->port);
+    }
 
+    if (channel->task->event.fd >= 0) {
+        channel->task->handler->close_connection(channel->task);
+    }
     __sync_bool_compare_and_swap(&channel->task->canceled, 1, 0);
     if ((result=sf_nio_notify(channel->task, SF_NIO_STAGE_CONNECT)) == 0) {
         channel->last_connect_time = g_current_time;
@@ -226,8 +254,8 @@ int idempotency_client_channel_check_reconnect(
 }
 
 struct idempotency_client_channel *idempotency_client_channel_get(
-        const char *server_ip, const uint16_t server_port,
-        const int timeout, int *err_no)
+        const FCCommunicationType comm_type, const char *server_ip,
+        const uint16_t server_port, const int timeout, int *err_no)
 {
     int r;
     int key_len;
@@ -239,8 +267,11 @@ struct idempotency_client_channel *idempotency_client_channel_get(
     IdempotencyClientChannel *current;
     IdempotencyClientChannel *channel;
 
-    key_len = snprintf(key, sizeof(key), "%s_%u", server_ip, server_port);
-    hash_code = simple_hash(key, key_len);
+    key_len = strlen(server_ip);
+    memcpy(key, server_ip, key_len);
+    *(key + key_len++) = '-';
+    key_len += fc_itoa(server_port, key + key_len);
+    hash_code = fc_simple_hash(key, key_len);
     bucket = channel_context.htable.buckets +
         hash_code % channel_context.htable.capacity;
     previous = NULL;
@@ -277,8 +308,8 @@ struct idempotency_client_channel *idempotency_client_channel_get(
             break;
         }
 
-        channel->task = alloc_channel_task(channel,
-                hash_code, server_ip, server_port, err_no);
+        channel->task = alloc_channel_task(channel, hash_code,
+                comm_type, server_ip, server_port, err_no);
         if (channel->task == NULL) {
             fast_mblock_free_object(&channel_context.
                     channel_allocator, channel);
@@ -323,8 +354,8 @@ int idempotency_client_channel_push(struct idempotency_client_channel *channel,
     receipt->req_id = req_id;
     fc_queue_push_ex(&channel->queue, receipt, &notify);
     if (notify) {
-        if (__sync_add_and_fetch(&channel->in_ioevent, 0)) {
-            if (__sync_add_and_fetch(&channel->established, 0)) {
+        if (FC_ATOMIC_GET(channel->in_ioevent)) {
+            if (FC_ATOMIC_GET(channel->established)) {
                 sf_nio_notify(channel->task, SF_NIO_STAGE_CONTINUE);
             }
         } else {

src/idempotency/client/client_channel.h

@@ -22,6 +22,7 @@
 #include "fastcommon/pthread_func.h"
 #include "fastcommon/sched_thread.h"
 #include "fastcommon/fc_atomic.h"
+#include "sf/sf_types.h"
 #include "client_types.h"
 
 #ifdef __cplusplus
@@ -40,13 +41,14 @@ void idempotency_client_channel_config_to_string_ex(
         char *output, const int size, const bool add_comma);
 
 struct idempotency_client_channel *idempotency_client_channel_get(
-        const char *server_ip, const uint16_t server_port,
-        const int timeout, int *err_no);
+        const FCCommunicationType comm_type, const char *server_ip,
+        const uint16_t server_port, const int timeout, int *err_no);
 
 static inline uint64_t idempotency_client_channel_next_seq_id(
         struct idempotency_client_channel *channel)
 {
-    return __sync_add_and_fetch(&channel->next_req_id, 1);
+    return SF_IDEMPOTENCY_NEXT_REQ_ID(channel->server_id,
+            channel->id, FC_ATOMIC_INC(channel->next_seq));
 }
 
 int idempotency_client_channel_push(struct idempotency_client_channel *channel,
@@ -74,13 +76,28 @@ static inline void idempotency_client_channel_set_id_key(
 static inline int idempotency_client_channel_check_wait_ex(
         struct idempotency_client_channel *channel, const int timeout)
 {
-    if (__sync_add_and_fetch(&channel->established, 0)) {
+    if (FC_ATOMIC_GET(channel->established)) {
         return 0;
     }
 
     idempotency_client_channel_check_reconnect(channel);
-    lcp_timedwait_sec(&channel->lc_pair, timeout);
-    return __sync_add_and_fetch(&channel->established, 0) ? 0 : ETIMEDOUT;
+    lcp_timedwait_sec(&channel->lcp, timeout);
+    if (FC_ATOMIC_GET(channel->established)) {
+        return 0;
+    } else {
+        /*
+        char formatted_ip[FORMATTED_IP_SIZE];
+        format_ip_address(channel->task->server_ip, formatted_ip);
+        logInfo("file: "__FILE__", line: %d, "
+                "channel_check fail, server %s:%u, in_ioevent: %d, "
+                "canceled: %d, req count: %"PRId64, __LINE__,
+                formatted_ip, channel->task->port,
+                __sync_add_and_fetch(&channel->in_ioevent, 0),
+                __sync_add_and_fetch(&channel->task->canceled, 0),
+                channel->task->req_count);
+         */
+        return ETIMEDOUT;
+    }
 }
 
 #ifdef __cplusplus

src/idempotency/client/client_types.h

@@ -21,6 +21,7 @@
 #include "fastcommon/fast_mblock.h"
 #include "fastcommon/fc_list.h"
 #include "fastcommon/fc_queue.h"
+#include "sf/idempotency/common/idempotency_types.h"
 
 typedef struct idempotency_client_config {
     bool enabled;
@@ -40,11 +41,12 @@ typedef struct idempotency_client_channel {
     volatile char in_ioevent;
     volatile char established;
     int buffer_size;  //the min task size of the server and mine
+    uint32_t server_id;
+    volatile uint32_t next_seq;
     time_t last_connect_time;  //for connect frequency control
     time_t last_pkg_time;      //last communication time
     time_t last_report_time;   //last report time for rpc receipt
-    pthread_lock_cond_pair_t lc_pair;  //for channel valid check and notify
-    volatile uint64_t next_req_id;
+    pthread_lock_cond_pair_t lcp;  //for channel valid check and notify
     struct fast_mblock_man receipt_allocator;
     struct fast_task_info *task;
     struct fc_queue queue;
@@ -61,6 +63,14 @@ typedef struct idempotency_receipt_thread_context {
     } last_check_times;
 } IdempotencyReceiptThreadContext;
 
+typedef struct idempotency_receipt_global_vars {
+    struct {
+        int task_padding_size;
+        sf_init_connection_callback init_connection;
+    } rdma;
+    IdempotencyReceiptThreadContext *thread_contexts;
+} IdempotencyReceiptGlobalVars;
+
 #ifdef __cplusplus
 extern "C" {
 #endif

src/idempotency/client/receipt_handler.c

@@ -41,23 +41,35 @@
 #include "client_channel.h"
 #include "receipt_handler.h"
 
-static IdempotencyReceiptThreadContext *receipt_thread_contexts = NULL;
+static IdempotencyReceiptGlobalVars receipt_global_vars;
 
-static int receipt_init_task(struct fast_task_info *task)
+#define RECEIPT_THREAD_CONTEXTS  receipt_global_vars.thread_contexts
+#define TASK_PADDING_SIZE        receipt_global_vars.rdma.task_padding_size
+#define RDMA_INIT_CONNECTION     receipt_global_vars.rdma.init_connection
+
+static int receipt_init_task(struct fast_task_info *task, void *arg)
 {
-    task->connect_timeout = SF_G_CONNECT_TIMEOUT; //for client side
-    task->network_timeout = SF_G_NETWORK_TIMEOUT;
-    return 0;
+#if IOEVENT_USE_URING
+    FC_URING_IS_CLIENT(task) = true;
+#endif
+
+    if (RDMA_INIT_CONNECTION != NULL) {
+        return RDMA_INIT_CONNECTION(task, arg);
+    } else {
+        return 0;
+    }
 }
 
 static int receipt_recv_timeout_callback(struct fast_task_info *task)
 {
     IdempotencyClientChannel *channel;
+    char formatted_ip[FORMATTED_IP_SIZE];
 
+    format_ip_address(task->server_ip, formatted_ip);
     if (SF_NIO_TASK_STAGE_FETCH(task) == SF_NIO_STAGE_CONNECT) {
         logError("file: "__FILE__", line: %d, "
                 "connect to server %s:%u timeout",
-                __LINE__, task->server_ip, task->port);
+                __LINE__, formatted_ip, task->port);
         return ETIMEDOUT;
     }
 
@@ -65,11 +77,13 @@ static int receipt_recv_timeout_callback(struct fast_task_info *task)
     if (channel->waiting_resp_qinfo.head != NULL) {
         logError("file: "__FILE__", line: %d, "
                 "waiting receipt response from server %s:%u timeout",
-                __LINE__, task->server_ip, task->port);
+                __LINE__, formatted_ip, task->port);
     } else {
         logError("file: "__FILE__", line: %d, "
-                "communication with server %s:%u timeout",
-                __LINE__, task->server_ip, task->port);
+                "%s server %s:%u timeout, channel established: %d",
+                __LINE__, task->nio_stages.current == SF_NIO_STAGE_SEND ?
+                "send to" : "recv from", formatted_ip, task->port,
+                FC_ATOMIC_GET(channel->established));
     }
 
     return ETIMEDOUT;
@@ -78,22 +92,27 @@ static int receipt_recv_timeout_callback(struct fast_task_info *task)
 static void receipt_task_finish_cleanup(struct fast_task_info *task)
 {
     IdempotencyClientChannel *channel;
+    char formatted_ip[FORMATTED_IP_SIZE];
 
     if (task->event.fd >= 0) {
         sf_task_detach_thread(task);
-        close(task->event.fd);
-        task->event.fd = -1;
     }
 
-    channel = (IdempotencyClientChannel *)task->arg;
+    sf_nio_reset_task_length(task);
+    task->req_count = 0;
+    task->pending_send_count = 0;
 
+    channel = (IdempotencyClientChannel *)task->arg;
     fc_list_del_init(&channel->dlink);
     __sync_bool_compare_and_swap(&channel->established, 1, 0);
     __sync_bool_compare_and_swap(&channel->in_ioevent, 1, 0);
 
-    logDebug("file: "__FILE__", line: %d, "
-            "receipt task for server %s:%u exit",
-            __LINE__, task->server_ip, task->port);
+    if (FC_LOG_BY_LEVEL(LOG_DEBUG)) {
+        format_ip_address(task->server_ip, formatted_ip);
+        logDebug("file: "__FILE__", line: %d, "
+                "receipt task for server %s:%u exit",
+                __LINE__, formatted_ip, task->port);
+    }
 }
 
 static void setup_channel_request(struct fast_task_info *task)
@@ -103,14 +122,15 @@ static void setup_channel_request(struct fast_task_info *task)
     SFProtoSetupChannelReq *req;
 
     channel = (IdempotencyClientChannel *)task->arg;
-    header = (SFCommonProtoHeader *)task->data;
+    header = (SFCommonProtoHeader *)task->send.ptr->data;
     req = (SFProtoSetupChannelReq *)(header + 1);
     int2buff(__sync_add_and_fetch(&channel->id, 0), req->channel_id);
     int2buff(__sync_add_and_fetch(&channel->key, 0), req->key);
 
     SF_PROTO_SET_HEADER(header, SF_SERVICE_PROTO_SETUP_CHANNEL_REQ,
             sizeof(SFProtoSetupChannelReq));
-    task->length = sizeof(SFCommonProtoHeader) + sizeof(SFProtoSetupChannelReq);
+    task->send.ptr->length = sizeof(SFCommonProtoHeader) +
+        sizeof(SFProtoSetupChannelReq);
     sf_send_add_event(task);
 }
 
@@ -131,15 +151,16 @@ static int check_report_req_receipt(struct fast_task_info *task)
         return 0;
     }
 
-    fc_queue_pop_to_queue(&channel->queue, &channel->waiting_resp_qinfo);
+    fc_queue_try_pop_to_queue(&channel->queue,
+            &channel->waiting_resp_qinfo);
     if (channel->waiting_resp_qinfo.head == NULL) {
         return 0;
     }
 
-    header = (SFCommonProtoHeader *)task->data;
+    header = (SFCommonProtoHeader *)task->send.ptr->data;
     rheader = (SFProtoReportReqReceiptHeader *)(header + 1);
     rbody = rstart = (SFProtoReportReqReceiptBody *)(rheader + 1);
-    buff_end = task->data + channel->buffer_size;
+    buff_end = task->send.ptr->data + channel->buffer_size;
     last = NULL;
     receipt = channel->waiting_resp_qinfo.head;
     do {
@@ -169,8 +190,9 @@ static int check_report_req_receipt(struct fast_task_info *task)
 
     count = rbody - rstart;
     int2buff(count, rheader->count);
-    task->length = (char *)rbody - task->data;
-    int2buff(task->length - sizeof(SFCommonProtoHeader), header->body_len);
+    task->send.ptr->length = (char *)rbody - task->send.ptr->data;
+    int2buff(task->send.ptr->length - sizeof(SFCommonProtoHeader),
+            header->body_len);
     header->cmd = SF_SERVICE_PROTO_REPORT_REQ_RECEIPT_REQ;
     sf_send_add_event(task);
     return count;
@@ -184,18 +206,18 @@ static void close_channel_request(struct fast_task_info *task)
     channel = (IdempotencyClientChannel *)task->arg;
     idempotency_client_channel_set_id_key(channel, 0, 0);
 
-    header = (SFCommonProtoHeader *)task->data;
+    header = (SFCommonProtoHeader *)task->send.ptr->data;
     SF_PROTO_SET_HEADER(header, SF_SERVICE_PROTO_CLOSE_CHANNEL_REQ, 0);
-    task->length = sizeof(SFCommonProtoHeader);
+    task->send.ptr->length = sizeof(SFCommonProtoHeader);
     sf_send_add_event(task);
 }
 
 static void active_test_request(struct fast_task_info *task)
 {
     SFCommonProtoHeader *header;
-    header = (SFCommonProtoHeader *)task->data;
+    header = (SFCommonProtoHeader *)task->send.ptr->data;
     SF_PROTO_SET_HEADER(header, SF_PROTO_ACTIVE_TEST_REQ, 0);
-    task->length = sizeof(SFCommonProtoHeader);
+    task->send.ptr->length = sizeof(SFCommonProtoHeader);
     sf_send_add_event(task);
 }
 
@@ -223,17 +245,22 @@ static void report_req_receipt_request(struct fast_task_info *task,
         if (update_lru) {
             update_lru_chain(task);
         }
+        task->pending_send_count++;
     }
 }
 
 static inline int receipt_expect_body_length(struct fast_task_info *task,
         const int expect_body_len)
 {
-    if ((int)(task->length - sizeof(SFCommonProtoHeader)) != expect_body_len) {
+    int body_len;
+    char formatted_ip[FORMATTED_IP_SIZE];
+
+    body_len = task->recv.ptr->length - sizeof(SFCommonProtoHeader);
+    if (body_len != expect_body_len) {
+        format_ip_address(task->server_ip, formatted_ip);
         logError("file: "__FILE__", line: %d, "
-                "server %s:%u, response body length: %d != %d",
-                __LINE__, task->server_ip, task->port, (int)(task->length -
-                    sizeof(SFCommonProtoHeader)), expect_body_len);
+                "server %s:%u, response body length: %d != %d", __LINE__,
+                formatted_ip, task->port, body_len, expect_body_len);
         return EINVAL;
     }
 
@@ -246,6 +273,7 @@ static int deal_setup_channel_response(struct fast_task_info *task)
     IdempotencyReceiptThreadContext *thread_ctx;
     SFProtoSetupChannelResp *resp;
     IdempotencyClientChannel *channel;
+    char formatted_ip[FORMATTED_IP_SIZE];
     int channel_id;
     int channel_key;
     int buffer_size;
@@ -257,28 +285,30 @@ static int deal_setup_channel_response(struct fast_task_info *task)
     }
 
     channel = (IdempotencyClientChannel *)task->arg;
-    if (__sync_add_and_fetch(&channel->established, 0)) {
+    if (FC_ATOMIC_GET(channel->established)) {
+        format_ip_address(task->server_ip, formatted_ip);
         logWarning("file: "__FILE__", line: %d, "
                 "response from server %s:%u, unexpected cmd: "
                 "SETUP_CHANNEL_RESP, ignore it!",
-                __LINE__, task->server_ip, task->port);
+                __LINE__, formatted_ip, task->port);
         return 0;
     }
 
-    resp = (SFProtoSetupChannelResp *)(task->data + sizeof(SFCommonProtoHeader));
+    resp = (SFProtoSetupChannelResp *)SF_PROTO_RECV_BODY(task);
     channel_id = buff2int(resp->channel_id);
     channel_key = buff2int(resp->key);
     buffer_size = buff2int(resp->buffer_size);
+    channel->server_id = buff2int(resp->server_id);
     idempotency_client_channel_set_id_key(channel, channel_id, channel_key);
     if (__sync_bool_compare_and_swap(&channel->established, 0, 1)) {
         thread_ctx = (IdempotencyReceiptThreadContext *)task->thread_data->arg;
         fc_list_add_tail(&channel->dlink, &thread_ctx->head);
     }
-    channel->buffer_size = FC_MIN(buffer_size, task->size);
+    channel->buffer_size = FC_MIN(buffer_size, task->send.ptr->size);
 
-    PTHREAD_MUTEX_LOCK(&channel->lc_pair.lock);
-    pthread_cond_broadcast(&channel->lc_pair.cond);
-    PTHREAD_MUTEX_UNLOCK(&channel->lc_pair.lock);
+    PTHREAD_MUTEX_LOCK(&channel->lcp.lock);
+    pthread_cond_broadcast(&channel->lcp.cond);
+    PTHREAD_MUTEX_UNLOCK(&channel->lcp.lock);
 
     if (channel->waiting_resp_qinfo.head != NULL) {
         bool notify;
@@ -297,6 +327,7 @@ static inline int deal_report_req_receipt_response(struct fast_task_info *task)
     IdempotencyClientChannel *channel;
     IdempotencyClientReceipt *current;
     IdempotencyClientReceipt *deleted;
+    char formatted_ip[FORMATTED_IP_SIZE];
 
     if ((result=receipt_expect_body_length(task, 0)) != 0) {
         return result;
@@ -304,13 +335,15 @@ static inline int deal_report_req_receipt_response(struct fast_task_info *task)
 
     channel = (IdempotencyClientChannel *)task->arg;
     if (channel->waiting_resp_qinfo.head == NULL) {
+        format_ip_address(task->server_ip, formatted_ip);
         logWarning("file: "__FILE__", line: %d, "
                 "response from server %s:%u, unexpect cmd: "
                 "REPORT_REQ_RECEIPT_RESP", __LINE__,
-                task->server_ip, task->port);
-        return 0;
+                formatted_ip, task->port);
+        return EINVAL;
     }
 
+    task->pending_send_count--;
     current = channel->waiting_resp_qinfo.head;
     do {
         deleted = current;
@@ -327,40 +360,46 @@ static inline int deal_report_req_receipt_response(struct fast_task_info *task)
 static int receipt_deal_task(struct fast_task_info *task, const int stage)
 {
     int result;
+    SFCommonProtoHeader *header;
+    char formatted_ip[FORMATTED_IP_SIZE];
 
     do {
         if (stage == SF_NIO_STAGE_HANDSHAKE) {
             setup_channel_request(task);
             result = 0;
             break;
-        } else if (stage == SF_NIO_STAGE_CONTINUE && task->length == 0) {
-            if (((IdempotencyClientChannel *)task->arg)->established) {
-                report_req_receipt_request(task, true);
-            } else {
-                sf_set_read_event(task);  //trigger read event
+        } else if (stage == SF_NIO_STAGE_CONTINUE) {
+            if (task->pending_send_count == 0) {
+                if (((IdempotencyClientChannel *)task->arg)->established) {
+                    report_req_receipt_request(task, true);
+                } else if (task->req_count > 0) {
+                    sf_set_read_event(task);  //trigger read event
+                }
             }
+
             result = 0;
             break;
         }
 
-        result = buff2short(((SFCommonProtoHeader *)task->data)->status);
+        header = (SFCommonProtoHeader *)task->recv.ptr->data;
+        result = buff2short(header->status);
         if (result != 0) {
             int msg_len;
             char *message;
 
-            msg_len = task->length - sizeof(SFCommonProtoHeader);
-            message = task->data + sizeof(SFCommonProtoHeader);
+            msg_len = SF_RECV_BODY_LENGTH(task);
+            message = SF_PROTO_RECV_BODY(task);
+            format_ip_address(task->server_ip, formatted_ip);
             logError("file: "__FILE__", line: %d, "
                     "response from server %s:%u, cmd: %d (%s), "
-                    "status: %d, error info: %.*s",
-                    __LINE__, task->server_ip, task->port,
-                    ((SFCommonProtoHeader *)task->data)->cmd,
-                    sf_get_cmd_caption(((SFCommonProtoHeader *)task->data)->cmd),
+                    "status: %d, error info: %.*s", __LINE__,
+                    formatted_ip, task->port, header->cmd,
+                    sf_get_cmd_caption(header->cmd),
                     result, msg_len, message);
             break;
         }
 
-        switch (((SFCommonProtoHeader *)task->data)->cmd) {
+        switch (header->cmd) {
             case SF_SERVICE_PROTO_SETUP_CHANNEL_RESP:
                 result = deal_setup_channel_response(task);
                 break;
@@ -368,28 +407,35 @@ static int receipt_deal_task(struct fast_task_info *task, const int stage)
                 result = deal_report_req_receipt_response(task);
                 break;
             case SF_PROTO_ACTIVE_TEST_RESP:
+                task->pending_send_count--;
                 result = 0;
                 break;
             case SF_SERVICE_PROTO_CLOSE_CHANNEL_RESP:
                 result = ECONNRESET; //force to close socket
-                logDebug("file: "__FILE__", line: %d, "
-                    "close channel to server %s:%u !!!",
-                    __LINE__, task->server_ip, task->port);
+                if (FC_LOG_BY_LEVEL(LOG_DEBUG)) {
+                    format_ip_address(task->server_ip, formatted_ip);
+                    logDebug("file: "__FILE__", line: %d, "
+                            "close channel to server %s:%u !!!",
+                            __LINE__, formatted_ip, task->port);
+                }
                 break;
             default:
+                format_ip_address(task->server_ip, formatted_ip);
                 logError("file: "__FILE__", line: %d, "
                         "response from server %s:%u, unexpect cmd: %d (%s)",
-                        __LINE__, task->server_ip, task->port,
-                        ((SFCommonProtoHeader *)task->data)->cmd,
-                        sf_get_cmd_caption(((SFCommonProtoHeader *)task->data)->cmd));
+                        __LINE__, formatted_ip, task->port, header->cmd,
+                        sf_get_cmd_caption(header->cmd));
                 result = EINVAL;
                 break;
         }
 
         if (result == 0) {
             update_lru_chain(task);
-            task->offset = task->length = 0;
-            report_req_receipt_request(task, false);
+            task->recv.ptr->length = 0;
+            task->recv.ptr->offset = 0;
+            if (task->pending_send_count == 0) {
+                report_req_receipt_request(task, false);
+            }
         }
     } while (0);
 
@@ -409,9 +455,10 @@ static void receipt_thread_check_heartbeat(
             break;
         }
 
-        if (sf_nio_task_is_idle(channel->task)) {
+        if (channel->task->pending_send_count == 0) {
             channel->last_pkg_time = g_current_time;
             active_test_request(channel->task);
+            channel->task->pending_send_count++;
         }
     }
 }
@@ -421,18 +468,22 @@ static void receipt_thread_close_idle_channel(
 {
     IdempotencyClientChannel *channel;
     IdempotencyClientChannel *tmp;
+    char formatted_ip[FORMATTED_IP_SIZE];
 
     fc_list_for_each_entry_safe(channel, tmp, &thread_ctx->head, dlink) {
-        if (!sf_nio_task_is_idle(channel->task)) {
+        if (channel->task->pending_send_count > 0) {
             continue;
         }
 
         if (g_current_time - channel->last_report_time >
                  g_idempotency_client_cfg.channel_max_idle_time)
         {
-            logDebug("file: "__FILE__", line: %d, "
-                    "close channel to server %s:%u because idle too long",
-                    __LINE__, channel->task->server_ip, channel->task->port);
+            if (FC_LOG_BY_LEVEL(LOG_DEBUG)) {
+                format_ip_address(channel->task->server_ip, formatted_ip);
+                logDebug("file: "__FILE__", line: %d, "
+                        "close channel to server %s:%u because idle too long",
+                        __LINE__, formatted_ip, channel->task->port);
+            }
             close_channel_request(channel->task);
         }
     }
@@ -463,40 +514,61 @@ static void *receipt_alloc_thread_extra_data(const int thread_index)
 {
     IdempotencyReceiptThreadContext *ctx;
 
-    ctx = receipt_thread_contexts + thread_index;
+    ctx = RECEIPT_THREAD_CONTEXTS + thread_index;
     FC_INIT_LIST_HEAD(&ctx->head);
     return ctx;
 }
 
-static int do_init()
+static int do_init(FCAddressPtrArray *address_array)
 {
+    const int task_arg_size = 0;
+    const bool double_buffers = false;
+    const bool need_shrink_task_buffer = false;
+    const bool explicit_post_recv = false;
+    int result;
     int bytes;
+    SFNetworkHandler *rdma_handler;
+    struct ibv_pd *pd;
 
     bytes = sizeof(IdempotencyReceiptThreadContext) * SF_G_WORK_THREADS;
-    receipt_thread_contexts = (IdempotencyReceiptThreadContext *)
+    RECEIPT_THREAD_CONTEXTS = (IdempotencyReceiptThreadContext *)
         fc_malloc(bytes);
-    if (receipt_thread_contexts == NULL) {
+    if (RECEIPT_THREAD_CONTEXTS == NULL) {
         return ENOMEM;
     }
-    memset(receipt_thread_contexts, 0, bytes);
+    memset(RECEIPT_THREAD_CONTEXTS, 0, bytes);
 
-    return sf_service_init_ex2(&g_sf_context,
+    if ((rdma_handler=sf_get_rdma_network_handler(&g_sf_context)) != NULL) {
+        if ((result=sf_alloc_rdma_pd(&g_sf_context, address_array)) != 0) {
+            return result;
+        }
+
+        TASK_PADDING_SIZE = rdma_handler->get_connection_size();
+        RDMA_INIT_CONNECTION = rdma_handler->init_connection;
+        pd = rdma_handler->pd;
+    } else {
+        TASK_PADDING_SIZE = 0;
+        RDMA_INIT_CONNECTION = NULL;
+        pd = NULL;
+    }
+    return sf_service_init_ex2(&g_sf_context, "idemp-receipt",
             receipt_alloc_thread_extra_data, receipt_thread_loop_callback,
-            NULL, sf_proto_set_body_length, receipt_deal_task,
+            NULL, sf_proto_set_body_length, NULL, NULL, receipt_deal_task,
             receipt_task_finish_cleanup, receipt_recv_timeout_callback,
-            1000, sizeof(SFCommonProtoHeader), 0, receipt_init_task);
+            1000, sizeof(SFCommonProtoHeader), TASK_PADDING_SIZE,
+            task_arg_size, double_buffers, need_shrink_task_buffer,
+            explicit_post_recv, receipt_init_task, pd, NULL);
 }
 
-int receipt_handler_init()
+int receipt_handler_init(FCAddressPtrArray *address_array)
 {
     int result;
 
-    if ((result=do_init()) != 0) {
+    if ((result=do_init(address_array)) != 0) {
         return result;
     }
 
     sf_enable_thread_notify(true);
-    sf_set_remove_from_ready_list(false);
     fc_sleep_ms(100);
 
     return 0;

src/idempotency/client/receipt_handler.h

@@ -24,7 +24,7 @@
 extern "C" {
 #endif
 
-int receipt_handler_init();
+int receipt_handler_init(FCAddressPtrArray *address_array);
 int receipt_handler_destroy();
 
 #ifdef __cplusplus

src/idempotency/client/rpc_wrapper.h

@@ -19,38 +19,39 @@
 
 #include "../../sf_configs.h"
 
-#define SF_CLIENT_IDEMPOTENCY_UPDATE_WRAPPER(client_ctx, \
+#define SF_CLIENT_IDEMPOTENCY_UPDATE_WRAPPER(client_ctx, conn_manager, \
         GET_MASTER_CONNECTION, get_conn_arg1, update_callback, ...) \
     ConnectionInfo *conn;  \
     IdempotencyClientChannel *old_channel;  \
     int result;  \
+    int conn_result;  \
     int i;       \
     bool idempotency_enabled;  \
     uint64_t req_id;  \
     SFNetRetryIntervalContext net_retry_ctx;  \
     \
-    if ((conn=GET_MASTER_CONNECTION(client_ctx, \
+    if ((conn=GET_MASTER_CONNECTION(conn_manager, \
                     get_conn_arg1, &result)) == NULL) \
     { \
         return SF_UNIX_ERRNO(result, EIO); \
     } \
-    connection_params = client_ctx->conn_manager. \
-            get_connection_params(client_ctx, conn); \
+    connection_params = (conn_manager)->ops. \
+            get_connection_params(conn_manager, conn); \
     idempotency_enabled = client_ctx->idempotency_enabled && \
             connection_params != NULL;  \
     \
     sf_init_net_retry_interval_context(&net_retry_ctx, \
-            &client_ctx->net_retry_cfg.interval_mm,    \
-            &client_ctx->net_retry_cfg.network);       \
+            &client_ctx->common_cfg.net_retry_cfg.interval_mm,    \
+            &client_ctx->common_cfg.net_retry_cfg.network);       \
+    \
+    if (idempotency_enabled) {   \
+        req_id = idempotency_client_channel_next_seq_id(  \
+                connection_params->channel); \
+    } else {  \
+        req_id = 0;  \
+    }  \
     \
     while (1) {  \
-        if (idempotency_enabled) {   \
-            req_id = idempotency_client_channel_next_seq_id(  \
-                    connection_params->channel); \
-        } else {  \
-            req_id = 0;  \
-        }  \
-    \
         old_channel = connection_params != NULL ? \
              connection_params->channel : NULL;   \
         i = 0; \
@@ -70,29 +71,40 @@
                 }  \
             }  \
     \
+            conn_result = result;  \
             if (result == SF_RETRIABLE_ERROR_CHANNEL_INVALID && \
                     idempotency_enabled)  \
             {  \
-                idempotency_client_channel_check_reconnect( \
-                        connection_params->channel);  \
+                if (idempotency_client_channel_check_wait(    \
+                            connection_params->channel) == 0) \
+                { \
+                    if ((conn_result=sf_proto_rebind_idempotency_channel( \
+                                conn, (conn_manager)->module_name, \
+                                connection_params->channel->id,  \
+                                connection_params->channel->key, \
+                                client_ctx->common_cfg.network_timeout)) == 0)   \
+                    { \
+                        continue; \
+                    } \
+                } \
             }  \
     \
             SF_NET_RETRY_CHECK_AND_SLEEP(net_retry_ctx, client_ctx-> \
-                    net_retry_cfg.network.times, ++i, result); \
+                    common_cfg.net_retry_cfg.network.times, ++i, result); \
     /* \
             logInfo("file: "__FILE__", line: %d, func: %s, "  \
                     "net retry result: %d, retry count: %d",  \
                     __LINE__, __FUNCTION__, result, i);       \
     */ \
-            SF_CLIENT_RELEASE_CONNECTION(client_ctx, conn, result); \
-            if ((conn=GET_MASTER_CONNECTION(client_ctx,  \
+            SF_CLIENT_RELEASE_CONNECTION(conn_manager, conn, conn_result); \
+            if ((conn=GET_MASTER_CONNECTION(conn_manager,  \
                             get_conn_arg1, &result)) == NULL)  \
             {  \
                 return SF_UNIX_ERRNO(result, EIO);  \
             }  \
     \
-            connection_params = client_ctx->conn_manager. \
-                get_connection_params(client_ctx, conn);  \
+            connection_params = (conn_manager)->ops. \
+                get_connection_params(conn_manager, conn);  \
             if (connection_params != NULL && connection_params->channel != \
                     old_channel) \
             {  \
@@ -106,33 +118,33 @@
             continue; \
         } \
     \
-        if (idempotency_enabled) { \
+        if (idempotency_enabled && !SF_IS_SERVER_RETRIABLE_ERROR(result)) { \
             idempotency_client_channel_push( \
                     connection_params->channel, req_id); \
         } \
         break; \
     } \
     \
-    SF_CLIENT_RELEASE_CONNECTION(client_ctx, conn, result); \
+    SF_CLIENT_RELEASE_CONNECTION(conn_manager, conn, result); \
     return SF_UNIX_ERRNO(result, EIO)
 
 
-#define SF_CLIENT_IDEMPOTENCY_QUERY_WRAPPER(client_ctx, \
+#define SF_CLIENT_IDEMPOTENCY_QUERY_WRAPPER(client_ctx, conn_manager, \
         GET_READABLE_CONNECTION, get_conn_arg1, query_callback, ...) \
     ConnectionInfo *conn;  \
     int result;  \
     int i;       \
     SFNetRetryIntervalContext net_retry_ctx;  \
     \
-    if ((conn=GET_READABLE_CONNECTION(client_ctx, \
+    if ((conn=GET_READABLE_CONNECTION(conn_manager, \
                     get_conn_arg1, &result)) == NULL) \
     { \
         return SF_UNIX_ERRNO(result, EIO); \
     } \
     \
     sf_init_net_retry_interval_context(&net_retry_ctx, \
-            &client_ctx->net_retry_cfg.interval_mm,    \
-            &client_ctx->net_retry_cfg.network);       \
+            &client_ctx->common_cfg.net_retry_cfg.interval_mm,    \
+            &client_ctx->common_cfg.net_retry_cfg.network);       \
     i = 0; \
     while (1) { \
         if ((result=query_callback(client_ctx,  \
@@ -141,21 +153,21 @@
             break;  \
         }  \
         SF_NET_RETRY_CHECK_AND_SLEEP(net_retry_ctx, client_ctx-> \
-                net_retry_cfg.network.times, ++i, result); \
+                common_cfg.net_retry_cfg.network.times, ++i, result); \
     /* \
         logInfo("file: "__FILE__", line: %d, func: %s, "  \
                 "net retry result: %d, retry count: %d",  \
                 __LINE__, __FUNCTION__, result, i);       \
     */ \
-        SF_CLIENT_RELEASE_CONNECTION(client_ctx, conn, result); \
-        if ((conn=GET_READABLE_CONNECTION(client_ctx,  \
+        SF_CLIENT_RELEASE_CONNECTION(conn_manager, conn, result); \
+        if ((conn=GET_READABLE_CONNECTION(conn_manager,  \
                         get_conn_arg1, &result)) == NULL)  \
         {  \
             return SF_UNIX_ERRNO(result, EIO);  \
         }  \
     } \
     \
-    SF_CLIENT_RELEASE_CONNECTION(client_ctx, conn, result); \
+    SF_CLIENT_RELEASE_CONNECTION(conn_manager, conn, result); \
     return SF_UNIX_ERRNO(result, EIO)
 
 

src/idempotency/common/idempotency_types.h

@@ -0,0 +1,46 @@
+/*
+ * Copyright (c) 2020 YuQing <384681@qq.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <https://www.gnu.org/licenses/>.
+ */
+
+
+#ifndef _IDEMPOTENCY_COMMON_TYPES_H
+#define _IDEMPOTENCY_COMMON_TYPES_H
+
+#include "fastcommon/common_define.h"
+
+#define SF_IDEMPOTENCY_CHANNEL_ID_BITS    16
+#define SF_IDEMPOTENCY_REQUEST_ID_BITS    (64 - SF_IDEMPOTENCY_CHANNEL_ID_BITS)
+#define SF_IDEMPOTENCY_MAX_CHANNEL_COUNT  ((1 << SF_IDEMPOTENCY_CHANNEL_ID_BITS) - 1)
+#define SF_IDEMPOTENCY_MAX_CHANNEL_ID     SF_IDEMPOTENCY_MAX_CHANNEL_COUNT
+
+#define SF_IDEMPOTENCY_SERVER_ID_OFFSET   48
+#define SF_IDEMPOTENCY_CHANNEL_ID_OFFSET  32
+
+#define SF_IDEMPOTENCY_NEXT_REQ_ID(server_id, channel_id, seq)    \
+    (((int64_t)server_id)  << SF_IDEMPOTENCY_SERVER_ID_OFFSET)  | \
+    (((int64_t)channel_id) << SF_IDEMPOTENCY_CHANNEL_ID_OFFSET) | \
+    (int64_t)seq
+
+#define SF_IDEMPOTENCY_EXTRACT_SERVER_ID(req_id) \
+    (int)((req_id >> SF_IDEMPOTENCY_SERVER_ID_OFFSET) & 0xFFFF)
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif

src/idempotency/server/request_metadata.c

@@ -0,0 +1,250 @@
+/*
+ * Copyright (c) 2020 YuQing <384681@qq.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <https://www.gnu.org/licenses/>.
+ */
+
+#include <limits.h>
+#include <sys/stat.h>
+#include "fastcommon/shared_func.h"
+#include "fastcommon/logger.h"
+#include "fastcommon/fc_atomic.h"
+#include "sf/sf_global.h"
+#include "request_metadata.h"
+
+static struct {
+    int process_interval_ms;
+    int master_side_timeout;  //in seconds
+    struct {
+        IdempotencyRequestMetadataContext *head;
+        IdempotencyRequestMetadataContext *tail;
+    } list;
+} g_request_metadata = {1000, 300, {NULL, NULL}};
+
+
+#define CHECK_MASTER_METADATA(meta) \
+    (meta != NULL && g_current_time - (long)meta->enqueue_time > \
+     g_request_metadata.master_side_timeout)
+
+static void process_master_side(IdempotencyRequestMetadataContext *ctx)
+{
+    struct fast_mblock_chain chain;
+    struct fast_mblock_node *node;
+
+    chain.head = chain.tail = NULL;
+    PTHREAD_MUTEX_LOCK(&ctx->lock);
+    if (CHECK_MASTER_METADATA(ctx->list.head)) {
+        do {
+            node = fast_mblock_to_node_ptr(ctx->list.head);
+            if (chain.head == NULL) {
+                chain.head = node;
+            } else {
+                chain.tail->next = node;
+            }
+            chain.tail = node;
+
+            ctx->list.head = ctx->list.head->next;
+        } while (CHECK_MASTER_METADATA(ctx->list.head));
+
+        if (ctx->list.head == NULL) {
+            ctx->list.tail = NULL;
+        }
+        chain.tail->next = NULL;
+    }
+
+    if (chain.head != NULL) {
+        fast_mblock_batch_free(&ctx->allocator, &chain);
+    }
+    PTHREAD_MUTEX_UNLOCK(&ctx->lock);
+}
+
+#define CHECK_SLAVE_METADATA(meta, dv) \
+    (meta != NULL && meta->data_version <= dv)
+
+static void process_slave_side(IdempotencyRequestMetadataContext *ctx,
+        const int64_t data_version)
+{
+    struct fast_mblock_chain chain;
+    struct fast_mblock_node *node;
+
+    chain.head = chain.tail = NULL;
+    PTHREAD_MUTEX_LOCK(&ctx->lock);
+    if (CHECK_SLAVE_METADATA(ctx->list.head, data_version)) {
+        do {
+            node = fast_mblock_to_node_ptr(ctx->list.head);
+            if (chain.head == NULL) {
+                chain.head = node;
+            } else {
+                chain.tail->next = node;
+            }
+            chain.tail = node;
+
+            ctx->list.head = ctx->list.head->next;
+        } while (CHECK_SLAVE_METADATA(ctx->list.head, data_version));
+
+        if (ctx->list.head == NULL) {
+            ctx->list.tail = NULL;
+        }
+        chain.tail->next = NULL;
+    }
+
+    if (chain.head != NULL) {
+        fast_mblock_batch_free(&ctx->allocator, &chain);
+    }
+    PTHREAD_MUTEX_UNLOCK(&ctx->lock);
+}
+
+static void *thread_run(void *arg)
+{
+    IdempotencyRequestMetadataContext *ctx;
+    int64_t data_version;
+
+#ifdef OS_LINUX
+    prctl(PR_SET_NAME, "idemp-req-meta");
+#endif
+
+    ctx = g_request_metadata.list.head;
+    while (SF_G_CONTINUE_FLAG) {
+        fc_sleep_ms(g_request_metadata.process_interval_ms);
+
+        if (ctx->is_master_callback.func(ctx->is_master_callback.
+                    arg, &data_version))
+        {
+            process_master_side(ctx);
+        } else if (data_version > 0) {
+            process_slave_side(ctx, data_version);
+        }
+
+        ctx = ctx->next;
+        if (ctx == NULL) {
+            ctx = g_request_metadata.list.head;
+        }
+    }
+
+    return NULL;
+}
+
+int idempotency_request_metadata_init(IdempotencyRequestMetadataContext
+        *ctx, sf_is_master_callback is_master_callback, void *arg)
+{
+    int result;
+
+    if ((result=fast_mblock_init_ex1(&ctx->allocator, "req-metadata-info",
+                    sizeof(IdempotencyRequestMetadata), 8192, 0,
+                    NULL, NULL, false)) != 0)
+    {
+        return result;
+    }
+
+    if ((result=init_pthread_lock(&ctx->lock)) != 0) {
+        return result;
+    }
+
+    ctx->is_master_callback.func = is_master_callback;
+    ctx->is_master_callback.arg = arg;
+    ctx->list.head = ctx->list.tail = NULL;
+
+    ctx->next = NULL;
+    if (g_request_metadata.list.head == NULL) {
+        g_request_metadata.list.head = ctx;
+    } else {
+        g_request_metadata.list.tail->next = ctx;
+    }
+    g_request_metadata.list.tail = ctx;
+
+    return 0;
+}
+
+int idempotency_request_metadata_start(const int process_interval_ms,
+            const int master_side_timeout)
+{
+    pthread_t tid;
+
+    if (g_request_metadata.list.head == NULL) {
+        logError("file: "__FILE__", line: %d, "
+                "list is empty!", __LINE__);
+        return ENOENT;
+    }
+
+    if (process_interval_ms <= 0) {
+        logError("file: "__FILE__", line: %d, "
+                "invalid process interval: %d!",
+                __LINE__, process_interval_ms);
+        return EINVAL;
+    }
+
+    if (master_side_timeout <= 0) {
+        logError("file: "__FILE__", line: %d, "
+                "invalid master side timeout: %d!",
+                __LINE__, master_side_timeout);
+        return EINVAL;
+    }
+
+    g_request_metadata.process_interval_ms = process_interval_ms;
+    g_request_metadata.master_side_timeout = master_side_timeout;
+    return fc_create_thread(&tid, thread_run, NULL,
+            SF_G_THREAD_STACK_SIZE);
+}
+
+int idempotency_request_metadata_add(IdempotencyRequestMetadataContext
+        *ctx, const SFRequestMetadata *metadata, const int n)
+{
+    IdempotencyRequestMetadata *idemp_meta;
+
+    PTHREAD_MUTEX_LOCK(&ctx->lock);
+    do {
+        if ((idemp_meta=fast_mblock_alloc_object(&ctx->allocator)) == NULL) {
+            break;
+        }
+
+        idemp_meta->req_id = metadata->req_id;
+        idemp_meta->data_version = metadata->data_version;
+        idemp_meta->n = n;
+        idemp_meta->enqueue_time = g_current_time;
+        idemp_meta->next = NULL;
+
+        if (ctx->list.head == NULL) {
+            ctx->list.head = idemp_meta;
+        } else {
+            ctx->list.tail->next = idemp_meta;
+        }
+        ctx->list.tail = idemp_meta;
+    } while (0);
+    PTHREAD_MUTEX_UNLOCK(&ctx->lock);
+
+    return (idemp_meta != NULL ? 0 : ENOMEM);
+}
+
+int idempotency_request_metadata_get(IdempotencyRequestMetadataContext
+        *ctx, const int64_t req_id, int64_t *data_version, int *n)
+{
+    int result;
+    IdempotencyRequestMetadata *meta;
+
+    result = ENOENT;
+    PTHREAD_MUTEX_LOCK(&ctx->lock);
+    meta = ctx->list.head;
+    while (meta != NULL) {
+        if (req_id == meta->req_id) {
+            result = 0;
+            *data_version = meta->data_version;
+            if (n != NULL) {
+                *n = meta->n;
+            }
+            break;
+        }
+        meta = meta->next;
+    }
+    PTHREAD_MUTEX_UNLOCK(&ctx->lock);
+
+    return result;
+}

src/idempotency/server/request_metadata.h

@@ -0,0 +1,66 @@
+/*
+ * Copyright (c) 2020 YuQing <384681@qq.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <https://www.gnu.org/licenses/>.
+ */
+
+
+#ifndef _SF_IDEMPOTENCY_REQUEST_METADATA_H
+#define _SF_IDEMPOTENCY_REQUEST_METADATA_H
+
+#include "server_types.h"
+
+typedef bool (*sf_is_master_callback)(void *arg, int64_t *data_version);
+
+typedef struct idempotency_request_metadata {
+    int64_t req_id;
+    int64_t data_version;
+    int n;  //integer argument
+    uint32_t enqueue_time;
+    struct idempotency_request_metadata *next;
+} IdempotencyRequestMetadata;
+
+typedef struct idempotency_request_metadata_context {
+    struct {
+        sf_is_master_callback func;
+        void *arg;
+    } is_master_callback;
+    struct fast_mblock_man allocator;  //element: IdempotencyRequestMetadata
+    pthread_mutex_t lock;
+    struct {
+        IdempotencyRequestMetadata *head;
+        IdempotencyRequestMetadata *tail;
+    } list;
+    struct idempotency_request_metadata_context *next;
+} IdempotencyRequestMetadataContext;
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+    int idempotency_request_metadata_init(IdempotencyRequestMetadataContext
+            *ctx, sf_is_master_callback is_master_callback, void *arg);
+
+    int idempotency_request_metadata_start(const int process_interval_ms,
+            const int master_side_timeout);
+
+    int idempotency_request_metadata_add(IdempotencyRequestMetadataContext
+            *ctx, const SFRequestMetadata *metadata, const int n);
+
+    int idempotency_request_metadata_get(IdempotencyRequestMetadataContext
+            *ctx, const int64_t req_id, int64_t *data_version, int *n);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif

src/idempotency/server/server_channel.c

@@ -100,7 +100,7 @@ int idempotency_channel_init(const uint32_t max_channel_id,
     element_size = sizeof(IdempotencyChannel) + sizeof(IdempotencyRequest *) *
         request_htable_capacity;
     if ((result=fast_mblock_init_ex1(&channel_context.channel_allocator,
-                    "channel_info", element_size, 1024, max_channel_id,
+                    "channel-info", element_size, 1024, max_channel_id,
                     idempotency_channel_alloc_init, NULL, true)) != 0)
     {
         return result;

src/idempotency/server/server_handler.c

@@ -37,12 +37,9 @@
 #include "server_channel.h"
 #include "server_handler.h"
 
-#define SF_TASK_BODY_LENGTH(task) \
-    (task->length - sizeof(SFCommonProtoHeader))
-
 int sf_server_deal_setup_channel(struct fast_task_info *task,
-        int *task_type, IdempotencyChannel **channel,
-        SFResponseInfo *response)
+        int *task_type, const int server_id, IdempotencyChannel
+        **channel, SFResponseInfo *response)
 {
     int result;
     SFProtoSetupChannelReq *req;
@@ -52,13 +49,13 @@ int sf_server_deal_setup_channel(struct fast_task_info *task,
 
     response->header.cmd = SF_SERVICE_PROTO_SETUP_CHANNEL_RESP;
     if ((result=sf_server_expect_body_length(response,
-                    SF_TASK_BODY_LENGTH(task),
+                    SF_RECV_BODY_LENGTH(task),
                     sizeof(SFProtoSetupChannelReq))) != 0)
     {
         return result;
     }
 
-    req = (SFProtoSetupChannelReq *)(task->data + sizeof(SFCommonProtoHeader));
+    req = (SFProtoSetupChannelReq *)SF_PROTO_RECV_BODY(task);
     channel_id = buff2int(req->channel_id);
     key = buff2int(req->key);
     if (*channel != NULL) {
@@ -74,14 +71,13 @@ int sf_server_deal_setup_channel(struct fast_task_info *task,
                 "alloc channel fail, hint channel id: %d", channel_id);
         return ENOMEM;
     }
-
     *task_type = SF_SERVER_TASK_TYPE_CHANNEL_HOLDER;
 
-    resp = (SFProtoSetupChannelResp *)(task->data +
-            sizeof(SFCommonProtoHeader));
+    resp = (SFProtoSetupChannelResp *)SF_PROTO_SEND_BODY(task);
     int2buff((*channel)->id, resp->channel_id);
     int2buff((*channel)->key, resp->key);
-    int2buff(task->size, resp->buffer_size);
+    int2buff(server_id, resp->server_id);
+    int2buff(task->send.ptr->size, resp->buffer_size);
     response->header.body_len = sizeof(SFProtoSetupChannelResp);
     return 0;
 }
@@ -135,19 +131,19 @@ int sf_server_deal_report_req_receipt(struct fast_task_info *task,
     SFProtoReportReqReceiptBody *body_part;
     SFProtoReportReqReceiptBody *body_end;
 
+    response->header.cmd = SF_SERVICE_PROTO_REPORT_REQ_RECEIPT_RESP;
     if ((result=check_holder_channel(task_type, channel, response)) != 0) {
         return result;
     }
 
-    body_len = SF_TASK_BODY_LENGTH(task);
+    body_len = SF_RECV_BODY_LENGTH(task);
     if ((result=sf_server_check_min_body_length(response, body_len,
                     sizeof(SFProtoReportReqReceiptHeader))) != 0)
     {
         return result;
     }
 
-    body_header = (SFProtoReportReqReceiptHeader *)
-    (task->data + sizeof(SFCommonProtoHeader));
+    body_header = (SFProtoReportReqReceiptHeader *)SF_PROTO_RECV_BODY(task);
     count = buff2int(body_header->count);
     calc_body_len = sizeof(SFProtoReportReqReceiptHeader) +
         sizeof(SFProtoReportReqReceiptBody) * count;
@@ -169,12 +165,11 @@ int sf_server_deal_report_req_receipt(struct fast_task_info *task,
     }
 
     //logInfo("receipt count: %d, success: %d", count, success);
-    response->header.cmd = SF_SERVICE_PROTO_REPORT_REQ_RECEIPT_RESP;
     return 0;
 }
 
 IdempotencyRequest *sf_server_update_prepare_and_check(
-        struct fast_task_info *task, struct fast_mblock_man *
+        SFRequestInfo *req, struct fast_mblock_man *
         request_allocator, IdempotencyChannel *channel,
         SFResponseInfo *response, int *result)
 {
@@ -188,9 +183,8 @@ IdempotencyRequest *sf_server_update_prepare_and_check(
         return NULL;
     }
 
-    adheader = (SFProtoIdempotencyAdditionalHeader *)
-        (task->data + sizeof(SFCommonProtoHeader));
-    request = (IdempotencyRequest *)fast_mblock_alloc_object(request_allocator);
+    adheader = (SFProtoIdempotencyAdditionalHeader *)req->body;
+    request = fast_mblock_alloc_object(request_allocator);
     if (request == NULL) {
         *result = ENOMEM;
         return NULL;
@@ -201,9 +195,58 @@ IdempotencyRequest *sf_server_update_prepare_and_check(
     *result = idempotency_channel_add_request(channel, request);
     if (*result == EEXIST) {
         if (!request->finished) {
+            response->error.length = sprintf(response->error.message,
+                    "idempotency req id: %"PRId64" exists but NOT "
+                    "finished", request->req_id);
             *result = EAGAIN;
         }
     }
 
     return request;
 }
+
+int sf_server_deal_rebind_channel(struct fast_task_info *task,
+        int *server_task_type, IdempotencyChannel **channel,
+        SFResponseInfo *response)
+{
+    int result;
+    uint32_t channel_id;
+    int key;
+    SFProtoRebindChannelReq *req;
+
+    if ((result=sf_server_expect_body_length(response,
+                    SF_RECV_BODY_LENGTH(task),
+                    sizeof(SFProtoRebindChannelReq))) != 0)
+    {
+        return result;
+    }
+
+    if (*server_task_type != SF_SERVER_TASK_TYPE_CHANNEL_USER) {
+        response->error.length = sprintf(response->error.message,
+                "invalid task type: %d != %d", *server_task_type,
+                SF_SERVER_TASK_TYPE_CHANNEL_USER);
+        return EINVAL;
+    }
+
+    if (*channel == NULL) {
+        response->error.length = sprintf(response->error.message,
+                "no channel binded");
+        return EINVAL;
+    }
+    idempotency_channel_release(*channel, false);
+
+    req = (SFProtoRebindChannelReq *)SF_PROTO_RECV_BODY(task);
+    channel_id = buff2int(req->channel_id);
+    key = buff2int(req->key);
+    *channel = idempotency_channel_find_and_hold(channel_id, key, &result);
+    if (*channel == NULL) {
+        response->error.length = sprintf(response->error.message,
+                "find channel fail, channel id: %d, result: %d",
+                channel_id, result);
+        *server_task_type = SF_SERVER_TASK_TYPE_NONE;
+        return SF_RETRIABLE_ERROR_NO_CHANNEL;
+    }
+
+    response->header.cmd = SF_SERVICE_PROTO_REBIND_CHANNEL_RESP;
+    return 0;
+}

src/idempotency/server/server_handler.h

@@ -25,8 +25,8 @@ extern "C" {
 #endif
 
 int sf_server_deal_setup_channel(struct fast_task_info *task,
-        int *task_type, IdempotencyChannel **channel,
-        SFResponseInfo *response);
+        int *task_type, const int server_id, IdempotencyChannel
+        **channel, SFResponseInfo *response);
 
 int sf_server_deal_close_channel(struct fast_task_info *task,
         int *task_type, IdempotencyChannel **channel,
@@ -37,10 +37,14 @@ int sf_server_deal_report_req_receipt(struct fast_task_info *task,
         SFResponseInfo *response);
 
 IdempotencyRequest *sf_server_update_prepare_and_check(
-        struct fast_task_info *task, struct fast_mblock_man *
+        SFRequestInfo *req, struct fast_mblock_man *
         request_allocator, IdempotencyChannel *channel,
         SFResponseInfo *response, int *result);
 
+int sf_server_deal_rebind_channel(struct fast_task_info *task,
+        int *server_task_type, IdempotencyChannel **channel,
+        SFResponseInfo *response);
+
 #ifdef __cplusplus
 }
 #endif

src/idempotency/server/server_types.h

@@ -19,11 +19,7 @@
 
 #include "fastcommon/fast_mblock.h"
 #include "fastcommon/fast_timer.h"
-
-#define SF_IDEMPOTENCY_CHANNEL_ID_BITS    16
-#define SF_IDEMPOTENCY_REQUEST_ID_BITS    (64 - SF_IDEMPOTENCY_CHANNEL_ID_BITS)
-#define SF_IDEMPOTENCY_MAX_CHANNEL_COUNT  ((1 << SF_IDEMPOTENCY_CHANNEL_ID_BITS) - 1)
-#define SF_IDEMPOTENCY_MAX_CHANNEL_ID     SF_IDEMPOTENCY_MAX_CHANNEL_COUNT
+#include "sf/idempotency/common/idempotency_types.h"
 
 #define SF_IDEMPOTENCY_DEFAULT_REQUEST_HINT_CAPACITY      1023
 #define SF_IDEMPOTENCY_DEFAULT_CHANNEL_RESERVE_INTERVAL    600
@@ -32,14 +28,14 @@
 typedef struct idempotency_request_result {
     short rsize;  //response size defined by application
     short flags;  //for application
-    int result;
-    void *response;
+    volatile int result;
+    void * volatile response;
 } IdempotencyRequestResult;
 
 typedef struct idempotency_request {
     uint64_t req_id;
     volatile int ref_count;
-    bool finished;
+    volatile char finished;
     IdempotencyRequestResult output;
     struct fast_mblock_man *allocator;  //for free
     struct idempotency_request *next;
@@ -61,6 +57,17 @@ typedef struct idempotency_channel {
     struct idempotency_channel *next;
 } IdempotencyChannel;
 
+typedef struct sf_request_metadata {
+    int64_t req_id;
+    int64_t data_version;
+} SFRequestMetadata;
+
+typedef struct sf_request_metadata_array {
+    SFRequestMetadata *elts;
+    int count;
+    int alloc;
+} SFRequestMetadataArray;
+
 #ifdef __cplusplus
 extern "C" {
 #endif

src/sf_binlog_index.c

@@ -0,0 +1,274 @@
+/*
+ * Copyright (c) 2020 YuQing <384681@qq.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <https://www.gnu.org/licenses/>.
+ */
+
+#include "fastcommon/logger.h"
+#include "fastcommon/shared_func.h"
+#include "fastcommon/fc_memory.h"
+#include "sf_binlog_index.h"
+
+#define SF_BINLOG_HEADER_FIELD_COUNT   2
+#define SF_BINLOG_HEADER_FIELD_INDEX_RECORD_COUNT  0
+#define SF_BINLOG_HEADER_FIELD_INDEX_LAST_VERSION  1
+
+void sf_binlog_index_init(SFBinlogIndexContext *ctx, const char *name,
+        const char *filename, const int record_max_size,
+        const int array_elt_size, pack_record_func pack_record,
+        unpack_record_func unpack_record)
+{
+    memset(ctx, 0, sizeof(SFBinlogIndexContext));
+    ctx->name = name;
+    ctx->filename = fc_strdup(filename);
+    ctx->record_max_size = record_max_size;
+    ctx->array_elt_size = array_elt_size;
+    ctx->pack_record = pack_record;
+    ctx->unpack_record = unpack_record;
+}
+
+static int parse_header(const string_t *line, int *record_count,
+        int64_t *last_version, char *error_info)
+{
+    int count;
+    char *endptr;
+    string_t cols[SF_BINLOG_HEADER_FIELD_COUNT];
+
+    count = split_string_ex(line, ' ', cols,
+            SF_BINLOG_HEADER_FIELD_COUNT, false);
+    if (count != SF_BINLOG_HEADER_FIELD_COUNT) {
+        sprintf(error_info, "field count: %d != %d",
+                count, SF_BINLOG_HEADER_FIELD_COUNT);
+        return EINVAL;
+    }
+
+    SF_BINLOG_PARSE_INT_SILENCE(*record_count, "record count",
+            SF_BINLOG_HEADER_FIELD_INDEX_RECORD_COUNT, ' ', 0);
+    SF_BINLOG_PARSE_INT_SILENCE(*last_version, "last version",
+            SF_BINLOG_HEADER_FIELD_INDEX_LAST_VERSION, '\n', 0);
+    return 0;
+}
+
+static int parse(SFBinlogIndexContext *ctx, const string_t *lines,
+        const int row_count)
+{
+    int result;
+    int record_count;
+    char error_info[256];
+    const string_t *line;
+    const string_t *end;
+    void *bindex;
+
+    if ((result=parse_header(lines, &record_count, &ctx->
+                    last_version, error_info)) != 0)
+    {
+        logError("file: "__FILE__", line: %d, "
+                "%s index file: %s, parse header fail, error info: %s",
+                __LINE__, ctx->name, ctx->filename, error_info);
+        return result;
+    }
+
+    if (row_count != record_count + 1) {
+        logError("file: "__FILE__", line: %d, "
+                "%s index file: %s, line count: %d != record count: "
+                "%d + 1", __LINE__, ctx->name, ctx->filename,
+                row_count, record_count + 1);
+        return EINVAL;
+    }
+
+    ctx->index_array.alloc = 64;
+    while (ctx->index_array.alloc < record_count) {
+        ctx->index_array.alloc *= 2;
+    }
+    ctx->index_array.indexes = fc_malloc(ctx->array_elt_size *
+            ctx->index_array.alloc);
+    if (ctx->index_array.indexes == NULL) {
+        return ENOMEM;
+    }
+
+    end = lines + row_count;
+    bindex = ctx->index_array.indexes;
+    for (line=lines+1; line<end; line++) {
+        if ((result=ctx->unpack_record(line, bindex, error_info)) != 0) {
+            logError("file: "__FILE__", line: %d, "
+                    "%s index file: %s, parse line #%d fail, error "
+                    "info: %s", __LINE__, ctx->name, ctx->filename,
+                    (int)(line - lines) + 1, error_info);
+            return result;
+        }
+
+        bindex = (char *)bindex + ctx->array_elt_size;
+    }
+
+    ctx->index_array.count = row_count - 1;
+    return 0;
+}
+
+static int load(SFBinlogIndexContext *ctx)
+{
+    int result;
+    int row_count;
+    int64_t file_size;
+    string_t context;
+    string_t *lines;
+
+    if ((result=getFileContent(ctx->filename, &context.str,
+                    &file_size)) != 0)
+    {
+        return result;
+    }
+
+    context.len = file_size;
+    row_count = getOccurCount(context.str, '\n');
+    lines = (string_t *)fc_malloc(sizeof(string_t) * row_count);
+    if (lines == NULL) {
+        free(context.str);
+        return ENOMEM;
+    }
+
+    row_count = split_string_ex(&context, '\n', lines, row_count, true);
+    if (row_count > 0) {
+        result = parse(ctx, lines, row_count);
+    }
+    free(lines);
+    free(context.str);
+    return result;
+}
+
+int sf_binlog_index_load(SFBinlogIndexContext *ctx)
+{
+    int result;
+
+    if (access(ctx->filename, F_OK) == 0) {
+        return load(ctx);
+    } else if (errno == ENOENT) {
+        return 0;
+    } else {
+        result = errno != 0 ? errno : EPERM;
+        logError("file: "__FILE__", line: %d, "
+                "access file %s fail, "
+                "errno: %d, error info: %s", __LINE__,
+                ctx->filename, result, STRERROR(result));
+        return result;
+    }
+}
+
+static int save(SFBinlogIndexContext *ctx, const char *filename)
+{
+    char buff[16 * 1024];
+    char *bend;
+    void *index;
+    char *p;
+    int fd;
+    int len;
+    int i;
+    int result;
+
+    if ((fd=open(filename, O_WRONLY | O_CREAT | O_TRUNC |
+                    O_CLOEXEC, 0644)) < 0)
+    {
+        result = errno != 0 ? errno : EIO;
+        logError("file: "__FILE__", line: %d, "
+                "open file %s fail, errno: %d, error info: %s",
+                __LINE__, filename, result, STRERROR(result));
+        return result;
+    }
+
+    result = 0;
+    p = buff;
+    bend = buff + sizeof(buff);
+
+    p += fc_itoa(ctx->index_array.count, p);
+    *p++ = ' ';
+    p += fc_itoa(ctx->last_version, p);
+    *p++ = '\n';
+
+    index = ctx->index_array.indexes;
+    for (i=0; i<ctx->index_array.count; i++) {
+        if (bend - p < ctx->record_max_size) {
+            len = p - buff;
+            if (fc_safe_write(fd, buff, len) != len) {
+                result = errno != 0 ? errno : EIO;
+                logError("file: "__FILE__", line: %d, "
+                        "write file %s fail, errno: %d, error info: %s",
+                        __LINE__, filename, result, STRERROR(result));
+                break;
+            }
+            p = buff;
+        }
+
+        p += ctx->pack_record(p, index);
+        index = (char *)index + ctx->array_elt_size;
+    }
+
+    if (result == 0) {
+        len = p - buff;
+        if (len > 0 && fc_safe_write(fd, buff, len) != len) {
+            result = errno != 0 ? errno : EIO;
+            logError("file: "__FILE__", line: %d, "
+                    "write file %s fail, errno: %d, error info: %s",
+                    __LINE__, filename, result, STRERROR(result));
+        }
+    }
+
+    close(fd);
+    return result;
+}
+
+int sf_binlog_index_save(SFBinlogIndexContext *ctx)
+{
+    int result;
+    char tmp_filename[PATH_MAX];
+
+    fc_combine_two_strings(ctx->filename, "tmp", '.', tmp_filename);
+    if ((result=save(ctx, tmp_filename)) != 0) {
+        return result;
+    }
+
+    if (rename(tmp_filename, ctx->filename) != 0) {
+        result = errno != 0 ? errno : EIO;
+        logError("file: "__FILE__", line: %d, "
+                "rename file \"%s\" to \"%s\" fail, "
+                "errno: %d, error info: %s",
+                __LINE__, tmp_filename, ctx->filename,
+                result, STRERROR(result));
+        return result;
+    }
+
+    return 0;
+}
+
+int sf_binlog_index_expand_array(SFBinlogIndexArray *array,
+        const int elt_size)
+{
+    int alloc;
+    void *indexes;
+
+    if (array->alloc == 0) {
+        alloc = 1024;
+    } else {
+        alloc = array->alloc * 2;
+    }
+    indexes = fc_malloc(elt_size * alloc);
+    if (indexes == NULL) {
+        return ENOMEM;
+    }
+
+    if (array->count > 0) {
+        memcpy(indexes, array->indexes, elt_size * array->count);
+        free(array->indexes);
+    }
+
+    array->indexes = indexes;
+    array->alloc = alloc;
+    return 0;
+}

src/sf_binlog_index.h

@@ -0,0 +1,100 @@
+/*
+ * Copyright (c) 2020 YuQing <384681@qq.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <https://www.gnu.org/licenses/>.
+ */
+
+//sf_binlog_index.h
+
+#ifndef _SF_BINLOG_INDEX_H_
+#define _SF_BINLOG_INDEX_H_
+
+#include "fastcommon/common_define.h"
+
+#define SF_BINLOG_PARSE_INT_SILENCE(var, caption, index, endchr, min_val) \
+    do {   \
+        var = strtoll(cols[index].str, &endptr, 10);  \
+        if (*endptr != endchr || var < min_val) {    \
+            sprintf(error_info, "invalid %s: %.*s",  \
+                    caption, cols[index].len, cols[index].str); \
+            return EINVAL;  \
+        }  \
+    } while (0)
+
+#define SF_BINLOG_PARSE_INT_SILENCE2(var, caption, index, echr1, echr2, min_val) \
+    do {   \
+        var = strtoll(cols[index].str, &endptr, 10);  \
+        if (!(*endptr == echr1 || *endptr == echr2) || (var < min_val)) { \
+            sprintf(error_info, "invalid %s: %.*s",  \
+                    caption, cols[index].len, cols[index].str); \
+            return EINVAL;  \
+        }  \
+    } while (0)
+
+
+typedef int (*pack_record_func)(char *buff, void *record);
+typedef int (*unpack_record_func)(const string_t *line,
+        void *record, char *error_info);
+
+typedef struct sf_binlog_index_array {
+    void *indexes;
+    int alloc;
+    int count;
+} SFBinlogIndexArray;
+
+typedef struct sf_binlog_index_context {
+    const char *name;
+    char *filename;
+    int record_max_size;
+    int array_elt_size;
+    pack_record_func pack_record;
+    unpack_record_func unpack_record;
+    SFBinlogIndexArray index_array;
+    int64_t last_version;
+} SFBinlogIndexContext;
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+void sf_binlog_index_init(SFBinlogIndexContext *ctx, const char *name,
+        const char *filename, const int record_max_size,
+        const int array_elt_size, pack_record_func pack_record,
+        unpack_record_func unpack_record);
+
+int sf_binlog_index_load(SFBinlogIndexContext *ctx);
+
+int sf_binlog_index_save(SFBinlogIndexContext *ctx);
+
+int sf_binlog_index_expand_array(SFBinlogIndexArray *array,
+        const int elt_size);
+
+static inline int sf_binlog_index_expand(SFBinlogIndexContext *ctx)
+{
+    return sf_binlog_index_expand_array(&ctx->
+            index_array, ctx->array_elt_size);
+}
+
+static inline void sf_binlog_index_free(SFBinlogIndexContext *ctx)
+{
+    if (ctx->index_array.indexes != NULL) {
+        free(ctx->index_array.indexes);
+        ctx->index_array.indexes = NULL;
+        ctx->index_array.alloc = ctx->index_array.count = 0;
+    }
+}
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif

src/sf_binlog_writer.h

@@ -19,24 +19,25 @@
 #define _SF_BINLOG_WRITER_H_
 
 #include "fastcommon/fc_queue.h"
+#include "fastcommon/fc_atomic.h"
 #include "sf_types.h"
+#include "sf_file_writer.h"
 
 #define SF_BINLOG_THREAD_ORDER_MODE_FIXED       0
 #define SF_BINLOG_THREAD_ORDER_MODE_VARY        1
 
-#define SF_BINLOG_THREAD_TYPE_ORDER_BY_NONE     0
-#define SF_BINLOG_THREAD_TYPE_ORDER_BY_VERSION  1
+#define SF_BINLOG_WRITER_TYPE_ORDER_BY_NONE     0
+#define SF_BINLOG_WRITER_TYPE_ORDER_BY_VERSION  1
 
-#define SF_BINLOG_BUFFER_TYPE_WRITE_TO_FILE     0  //default type, must be 0
-#define SF_BINLOG_BUFFER_TYPE_SET_NEXT_VERSION  1
-
-#define SF_BINLOG_SUBDIR_NAME_SIZE 128
-#define SF_BINLOG_FILE_MAX_SIZE   (1024 * 1024 * 1024)  //for binlog rotating by size
-#define SF_BINLOG_FILE_PREFIX     "binlog"
-#define SF_BINLOG_FILE_EXT_FMT    ".%06d"
-
-#define SF_BINLOG_BUFFER_LENGTH(buffer) ((buffer).end - (buffer).buff)
-#define SF_BINLOG_BUFFER_REMAIN(buffer) ((buffer).end - (buffer).current)
+#define SF_BINLOG_BUFFER_TYPE_WRITE_TO_FILE        0  //default type, must be 0
+#define SF_BINLOG_BUFFER_TYPE_SET_NEXT_VERSION     1
+#define SF_BINLOG_BUFFER_TYPE_CHANGE_ORDER_TYPE    2
+#define SF_BINLOG_BUFFER_TYPE_CHANGE_PASSIVE_WRITE 3
+#define SF_BINLOG_BUFFER_TYPE_CHANGE_CALL_FSYNC    4
+#define SF_BINLOG_BUFFER_TYPE_SET_WRITE_INDEX      5
+#define SF_BINLOG_BUFFER_TYPE_ROTATE_FILE          6
+#define SF_BINLOG_BUFFER_TYPE_NOTIFY_EXIT          7
+#define SF_BINLOG_BUFFER_TYPE_FLUSH_FILE           8
 
 #define SF_BINLOG_BUFFER_SET_VERSION(buffer, ver)  \
     (buffer)->version.first = (buffer)->version.last = ver
@@ -46,27 +47,38 @@ struct sf_binlog_writer_info;
 typedef struct sf_binlog_writer_buffer {
     SFVersionRange version;
     BufferInfo bf;
-    int type;    //for versioned writer
+    int type;
+    uint32_t timestamp;  //for flow ctrol
     struct sf_binlog_writer_info *writer;
     struct sf_binlog_writer_buffer *next;
 } SFBinlogWriterBuffer;
 
+typedef struct sf_binlog_writer_slot {
+    SFBinlogWriterBuffer head;
+} SFBinlogWriterSlot;
+
 typedef struct sf_binlog_writer_buffer_ring {
-    SFBinlogWriterBuffer **entries;
-    SFBinlogWriterBuffer **start; //for consumer
-    SFBinlogWriterBuffer **end;   //for producer
-    int count;
-    int max_count;
+    SFBinlogWriterSlot *slots;
+    int waiting_count;
+    int max_waitings;
     int size;
 } SFBinlogWriterBufferRing;
 
 typedef struct binlog_writer_thread {
     struct fast_mblock_man mblock;
     struct fc_queue queue;
-    bool running;
+    char name[64];
+    volatile bool running;
     bool use_fixed_buffer_size;
-    short order_mode;
-    short order_by;
+    bool passive_write;
+    char order_mode;
+    int write_interval_ms;
+    struct {
+        int max_delay;  //in seconds
+        volatile uint32_t last_timestamp;
+        int waiting_count;
+        pthread_lock_cond_pair_t lcp;
+    } flow_ctrol;
     struct {
         struct sf_binlog_writer_info *head;
         struct sf_binlog_writer_info *tail;
@@ -74,28 +86,16 @@ typedef struct binlog_writer_thread {
 } SFBinlogWriterThread;
 
 typedef struct sf_binlog_writer_info {
-    struct {
-        char subdir_name[SF_BINLOG_SUBDIR_NAME_SIZE];
-        int max_record_size;
-    } cfg;
-
-    struct {
-        int index;
-        int compress_index;
-    } binlog;
-
-    struct {
-        int fd;
-        int64_t size;
-        char *name;
-    } file;
+    SFFileWriterInfo fw;
 
     struct {
         SFBinlogWriterBufferRing ring;
-        int64_t next;
+        volatile int64_t next;
+        int64_t change_count;  //version change count
     } version_ctx;
-    SFBinlogBuffer binlog_buffer;
     SFBinlogWriterThread *thread;
+
+    short order_by;
     struct {
         bool in_queue;
         struct sf_binlog_writer_info *next;
@@ -111,58 +111,187 @@ typedef struct sf_binlog_writer_context {
 extern "C" {
 #endif
 
-    extern char *g_sf_binlog_data_path;
+int sf_binlog_writer_init_normal_ex(SFBinlogWriterInfo *writer,
+        const char *data_path, const char *subdir_name,
+        const char *file_prefix, const int max_record_size,
+        const int buffer_size, const int64_t file_rotate_size,
+        const bool call_fsync);
 
-int sf_binlog_writer_init_normal(SFBinlogWriterInfo *writer,
-        const char *subdir_name, const int buffer_size);
-
-int sf_binlog_writer_init_by_version(SFBinlogWriterInfo *writer,
-        const char *subdir_name, const uint64_t next_version,
-        const int buffer_size, const int ring_size);
+int sf_binlog_writer_init_by_version_ex(SFBinlogWriterInfo *writer,
+        const char *data_path, const char *subdir_name,
+        const char *file_prefix, const int max_record_size,
+        const uint64_t next_version, const int buffer_size,
+        const int ring_size, const int64_t file_rotate_size,
+        const bool call_fsync);
 
 int sf_binlog_writer_init_thread_ex(SFBinlogWriterThread *thread,
-        SFBinlogWriterInfo *writer, const short order_mode,
-        const short order_by, const int max_record_size,
-        const int writer_count, const bool use_fixed_buffer_size);
+        const char *name, SFBinlogWriterInfo *writer, const short order_mode,
+        const int write_interval_ms, const int max_delay,
+        const int max_record_size, const bool use_fixed_buffer_size,
+        const bool passive_write);
 
-#define sf_binlog_writer_init_thread(thread, \
-        writer, order_by, max_record_size)   \
-    sf_binlog_writer_init_thread_ex(thread, writer, \
-            SF_BINLOG_THREAD_ORDER_MODE_FIXED,      \
-            order_by, max_record_size, 1, true)
+#define sf_binlog_writer_init_normal(writer, data_path, \
+        subdir_name, max_record_size, buffer_size)      \
+    sf_binlog_writer_init_normal_ex(writer, data_path, subdir_name,  \
+            SF_BINLOG_FILE_PREFIX_STR, max_record_size, buffer_size, \
+            SF_BINLOG_DEFAULT_ROTATE_SIZE, true)
 
-static inline int sf_binlog_writer_init(SFBinlogWriterContext *context,
-        const char *subdir_name, const int buffer_size,
-        const int max_record_size)
+#define sf_binlog_writer_init_by_version(writer, data_path, subdir_name, \
+        max_record_size, next_version, buffer_size, ring_size)           \
+    sf_binlog_writer_init_by_version_ex(writer, data_path, subdir_name,  \
+            SF_BINLOG_FILE_PREFIX_STR, max_record_size, next_version,    \
+            buffer_size, ring_size, SF_BINLOG_DEFAULT_ROTATE_SIZE, true)
+
+#define sf_binlog_writer_init_thread(thread, name, \
+        writer, write_interval_ms, max_delay, max_record_size)    \
+    sf_binlog_writer_init_thread_ex(thread, name, writer,         \
+            SF_BINLOG_THREAD_ORDER_MODE_FIXED, write_interval_ms, \
+            max_delay, max_record_size, true, false)
+
+static inline int sf_binlog_writer_init_ex(SFBinlogWriterContext *context,
+        const char *data_path, const char *subdir_name,
+        const char *file_prefix, const int buffer_size,
+        const int write_interval_ms, const int max_delay,
+        const int max_record_size, const bool call_fsync)
 {
     int result;
-    if ((result=sf_binlog_writer_init_normal(&context->writer,
-                    subdir_name, buffer_size)) != 0)
+    if ((result=sf_binlog_writer_init_normal_ex(&context->writer, data_path,
+                    subdir_name, file_prefix, max_record_size, buffer_size,
+                    SF_BINLOG_DEFAULT_ROTATE_SIZE, call_fsync)) != 0)
     {
         return result;
     }
 
-    return sf_binlog_writer_init_thread(&context->thread, &context->writer,
-            SF_BINLOG_THREAD_TYPE_ORDER_BY_NONE, max_record_size);
+    return sf_binlog_writer_init_thread(&context->thread, subdir_name,
+            &context->writer, write_interval_ms, max_delay, max_record_size);
 }
 
-int sf_binlog_writer_change_order_by(SFBinlogWriterThread *thread,
+#define sf_binlog_writer_init(context, data_path, subdir_name,      \
+        buffer_size, write_interval_ms, max_delay, max_record_size) \
+    sf_binlog_writer_init_ex(context, data_path, subdir_name,       \
+            SF_BINLOG_FILE_PREFIX_STR, buffer_size, write_interval_ms,  \
+            max_delay, max_record_size, true)
+
+void sf_binlog_writer_finish(SFBinlogWriterInfo *writer);
+
+static inline void sf_binlog_writer_destroy_writer(
+        SFBinlogWriterInfo *writer)
+{
+    sf_file_writer_destroy(&writer->fw);
+    if (writer->version_ctx.ring.slots != NULL) {
+        free(writer->version_ctx.ring.slots);
+        writer->version_ctx.ring.slots = NULL;
+    }
+}
+
+static inline void sf_binlog_writer_destroy_thread(
+        SFBinlogWriterThread *thread)
+{
+    fast_mblock_destroy(&thread->mblock);
+    fc_queue_destroy(&thread->queue);
+}
+
+static inline void sf_binlog_writer_destroy(
+        SFBinlogWriterContext *context)
+{
+    sf_binlog_writer_finish(&context->writer);
+    sf_binlog_writer_destroy_writer(&context->writer);
+    sf_binlog_writer_destroy_thread(&context->thread);
+}
+
+int sf_binlog_writer_change_order_by(SFBinlogWriterInfo *writer,
         const short order_by);
 
+int sf_binlog_writer_change_passive_write(SFBinlogWriterInfo *writer,
+        const bool passive_write);
+
+int sf_binlog_writer_change_call_fsync(SFBinlogWriterInfo *writer,
+        const bool call_fsync);
+
 int sf_binlog_writer_change_next_version(SFBinlogWriterInfo *writer,
         const int64_t next_version);
 
-void sf_binlog_writer_finish(SFBinlogWriterInfo *writer);
+static inline int64_t sf_binlog_writer_get_next_version(
+        SFBinlogWriterInfo *writer)
+{
+    return writer->version_ctx.next;
+}
 
-int sf_binlog_get_current_write_index(SFBinlogWriterInfo *writer);
+static inline int sf_binlog_writer_get_waiting_count(
+        SFBinlogWriterInfo *writer)
+{
+    return writer->version_ctx.ring.waiting_count;
+}
 
-void sf_binlog_get_current_write_position(SFBinlogWriterInfo *writer,
-        SFBinlogFilePosition *position);
+static inline int sf_binlog_writer_get_thread_waiting_count(
+        SFBinlogWriterThread *thread)
+{
+    int waiting_count;
+
+    PTHREAD_MUTEX_LOCK(&thread->flow_ctrol.lcp.lock);
+    waiting_count = thread->flow_ctrol.waiting_count;
+    PTHREAD_MUTEX_UNLOCK(&thread->flow_ctrol.lcp.lock);
+
+    return waiting_count;
+}
+
+int sf_binlog_writer_rotate_file_ex(SFBinlogWriterInfo *writer,
+        const bool skip_empty_file);
+
+static inline int sf_binlog_writer_rotate_file(SFBinlogWriterInfo *writer)
+{
+    const bool skip_empty_file = false;
+    return sf_binlog_writer_rotate_file_ex(writer, skip_empty_file);
+}
+
+int sf_binlog_writer_flush_file(SFBinlogWriterInfo *writer);
+
+int sf_binlog_writer_change_write_index(SFBinlogWriterInfo *writer,
+        const int write_index);
+
+int sf_binlog_writer_notify_exit(SFBinlogWriterInfo *writer);
+
+#define sf_binlog_writer_set_flags(writer, flags) \
+    sf_file_writer_set_flags(&(writer)->fw, flags)
+
+#define sf_binlog_writer_set_write_done_callback(writer, callback, args) \
+    sf_file_writer_set_write_done_callback(&(writer)->fw, callback, args)
+
+#define sf_binlog_writer_get_last_version_ex(writer, log_level) \
+    sf_file_writer_get_last_version_ex(&(writer)->fw, log_level)
+
+#define sf_binlog_writer_get_last_version(writer) \
+    sf_file_writer_get_last_version(&(writer)->fw)
+
+#define sf_binlog_writer_get_last_version_silence(writer) \
+    sf_file_writer_get_last_version_silence(&(writer)->fw)
+
+#define sf_binlog_get_indexes(writer, start_index, last_index) \
+    sf_file_writer_get_indexes(&(writer)->fw, start_index, last_index)
+
+#define sf_binlog_get_start_index(writer) \
+    sf_file_writer_get_start_index(&(writer)->fw)
+
+#define sf_binlog_get_last_index(writer) \
+    sf_file_writer_get_last_index(&(writer)->fw)
+
+#define sf_binlog_get_current_write_index(writer) \
+    sf_file_writer_get_current_write_index(&(writer)->fw)
+
+#define sf_binlog_get_current_write_position(writer, position) \
+    sf_file_writer_get_current_position(&(writer)->fw, position)
 
 static inline SFBinlogWriterBuffer *sf_binlog_writer_alloc_buffer(
         SFBinlogWriterThread *thread)
 {
-    return (SFBinlogWriterBuffer *)fast_mblock_alloc_object(&thread->mblock);
+    SFBinlogWriterBuffer *buffer;
+
+    if ((buffer=(SFBinlogWriterBuffer *)fast_mblock_alloc_object(
+                    &thread->mblock)) != NULL)
+    {
+        buffer->type = SF_BINLOG_BUFFER_TYPE_WRITE_TO_FILE;
+    }
+    return buffer;
 }
 
 #define sf_binlog_writer_alloc_one_version_buffer(writer, version) \
@@ -179,6 +308,7 @@ static inline SFBinlogWriterBuffer *sf_binlog_writer_alloc_versioned_buffer_ex(
         const int64_t last_version, const int type)
 {
     SFBinlogWriterBuffer *buffer;
+
     buffer = (SFBinlogWriterBuffer *)fast_mblock_alloc_object(
             &writer->thread->mblock);
     if (buffer != NULL) {
@@ -190,38 +320,55 @@ static inline SFBinlogWriterBuffer *sf_binlog_writer_alloc_versioned_buffer_ex(
     return buffer;
 }
 
-static inline const char *sf_binlog_writer_get_filepath(const char *subdir_name,
-        char *filename, const int size)
-{
-    snprintf(filename, size, "%s/%s", g_sf_binlog_data_path, subdir_name);
-    return filename;
-}
+void sf_push_to_binlog_write_queue(SFBinlogWriterInfo *writer,
+        SFBinlogWriterBuffer *buffer);
 
-static inline const char *sf_binlog_writer_get_filename(const char *subdir_name,
-        const int binlog_index, char *filename, const int size)
-{
-    snprintf(filename, size, "%s/%s/%s"SF_BINLOG_FILE_EXT_FMT,
-            g_sf_binlog_data_path, subdir_name,
-            SF_BINLOG_FILE_PREFIX, binlog_index);
-    return filename;
-}
+#define sf_binlog_writer_get_filepath(data_path, subdir_name, filepath, size) \
+    sf_file_writer_get_filepath(data_path, subdir_name, filepath, size)
 
-int sf_binlog_writer_set_binlog_index(SFBinlogWriterInfo *writer,
-        const int binlog_index);
+#define sf_binlog_writer_get_filename_ex(data_path, subdir_name, \
+        file_prefix, binlog_index, filename, size) \
+        sf_file_writer_get_filename_ex(data_path, subdir_name, \
+                file_prefix, binlog_index, filename, size)
 
-#define sf_push_to_binlog_thread_queue(thread, buffer) \
-    fc_queue_push(&(thread)->queue, buffer)
+#define sf_binlog_writer_get_filename(data_path, \
+        subdir_name, binlog_index, filename, size) \
+        sf_file_writer_get_filename(data_path, subdir_name, \
+                binlog_index, filename, size)
 
-static inline void sf_push_to_binlog_write_queue(SFBinlogWriterInfo *writer,
-        SFBinlogWriterBuffer *buffer)
-{
-    buffer->type = SF_BINLOG_BUFFER_TYPE_WRITE_TO_FILE;
-    fc_queue_push(&writer->thread->queue, buffer);
-}
+#define sf_binlog_writer_get_index_filename(data_path, \
+        subdir_name, filename, size) \
+        sf_file_writer_get_index_filename(data_path, \
+                subdir_name, filename, size)
 
-int sf_binlog_writer_get_last_lines(const char *subdir_name,
-        const int current_write_index, char *buff,
-        const int buff_size, int *count, int *length);
+#define sf_binlog_writer_get_binlog_indexes(data_path, \
+        subdir_name, start_index, last_index) \
+        sf_file_writer_get_binlog_indexes(data_path, \
+                subdir_name, start_index, last_index)
+
+#define sf_binlog_writer_get_binlog_start_index(data_path, \
+        subdir_name, start_index) \
+        sf_file_writer_get_binlog_start_index(data_path, \
+                subdir_name, start_index)
+
+#define sf_binlog_writer_get_binlog_last_index(data_path, \
+        subdir_name, last_index) \
+        sf_file_writer_get_binlog_last_index(data_path, \
+                subdir_name, last_index)
+
+#define sf_binlog_set_indexes(writer, start_index, last_index) \
+    sf_file_writer_set_indexes(&(writer)->fw, start_index, last_index)
+
+#define sf_binlog_writer_set_binlog_start_index(writer, start_index) \
+    sf_file_writer_set_binlog_start_index(&(writer)->fw, start_index)
+
+#define sf_binlog_writer_set_binlog_write_index(writer, last_index) \
+    sf_file_writer_set_binlog_write_index(&(writer)->fw, last_index)
+
+#define sf_binlog_writer_get_last_lines(data_path, subdir_name, \
+        current_write_index, buff, buff_size, count, length)  \
+        sf_file_writer_get_last_lines(data_path, subdir_name, \
+                current_write_index, buff, buff_size, count, length)
 
 #ifdef __cplusplus
 }

src/sf_buffered_writer.h

@@ -0,0 +1,106 @@
+/*
+ * Copyright (c) 2020 YuQing <384681@qq.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <https://www.gnu.org/licenses/>.
+ */
+
+//sf_buffered_writer.h
+
+#ifndef _SF_BUFFERED_WRITER_H_
+#define _SF_BUFFERED_WRITER_H_
+
+#include "sf_types.h"
+#include "sf_func.h"
+
+typedef struct {
+    int fd;
+    const char *filename;
+    SFBinlogBuffer buffer;
+} SFBufferedWriter;
+
+#define sf_buffered_writer_init(writer, filename) \
+    sf_buffered_writer_init_ex(writer, filename, 1024 * 1024)
+
+#define SF_BUFFERED_WRITER_LENGTH(bw)  \
+    SF_BINLOG_BUFFER_PRODUCER_DATA_LENGTH((bw).buffer)
+
+#define SF_BUFFERED_WRITER_REMAIN(bw)  \
+    SF_BINLOG_BUFFER_PRODUCER_BUFF_REMAIN((bw).buffer)
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+    static inline int sf_buffered_writer_init_ex(SFBufferedWriter *writer,
+            const char *filename, const int buffer_size)
+    {
+        int result;
+
+        writer->filename = filename;
+        writer->fd = open(filename, O_WRONLY | O_CREAT |
+                O_TRUNC | O_CLOEXEC, 0644);
+        if (writer->fd < 0) {
+            result = errno != 0 ? errno : EIO;
+            logError("file: "__FILE__", line: %d, "
+                    "open file %s fail, errno: %d, error info: %s",
+                    __LINE__, filename, result, STRERROR(result));
+            return result;
+        }
+
+        if ((result=sf_binlog_buffer_init(&writer->buffer, buffer_size)) != 0) {
+            return result;
+        }
+        return 0;
+    }
+
+    static inline int sf_buffered_writer_save(SFBufferedWriter *writer)
+    {
+        int result;
+        int length;
+
+        length = writer->buffer.data_end - writer->buffer.buff;
+        if (fc_safe_write(writer->fd, writer->buffer.buff, length) != length) {
+            result = errno != 0 ? errno : EIO;
+            logError("file: "__FILE__", line: %d, "
+                    "write to file %s fail, errno: %d, error info: %s",
+                    __LINE__, writer->filename, result, STRERROR(result));
+            return result;
+        }
+
+        writer->buffer.data_end = writer->buffer.buff;
+        return 0;
+    }
+
+    static inline int sf_buffered_writer_destroy(SFBufferedWriter *writer)
+    {
+        int result;
+
+        if (writer->fd >= 0) {
+            if (fsync(writer->fd) != 0) {
+                result = errno != 0 ? errno : EIO;
+                logError("file: "__FILE__", line: %d, "
+                        "fsync to file %s fail, errno: %d, error info: %s",
+                        __LINE__, writer->filename, result, STRERROR(result));
+                return result;
+            }
+            close(writer->fd);
+            writer->fd = -1;
+        }
+        sf_binlog_buffer_destroy(&writer->buffer);
+        return 0;
+    }
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif

src/sf_cluster_cfg.c

@@ -0,0 +1,139 @@
+/*
+ * Copyright (c) 2020 YuQing <384681@qq.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <https://www.gnu.org/licenses/>.
+ */
+
+#include <limits.h>
+#include "fastcommon/fast_buffer.h"
+#include "fastcommon/md5.h"
+#include "sf_cluster_cfg.h"
+
+static int calc_cluster_config_sign(SFClusterConfig *cluster)
+{
+    FastBuffer buffer;
+    int result;
+
+    if ((result=fast_buffer_init1(&buffer, 1024)) != 0) {
+        return result;
+    }
+    fc_server_to_config_string(&cluster->server_cfg, &buffer);
+    my_md5_buffer(buffer.data, buffer.length, cluster->md5_digest);
+
+    /*
+    {
+    char hex_buff[2 * sizeof(cluster->md5_digest) + 1];
+    logInfo("cluster config length: %d, sign: %s", buffer.length,
+            bin2hex((const char *)cluster->md5_digest,
+                sizeof(cluster->md5_digest), hex_buff));
+    }
+    */
+
+    fast_buffer_destroy(&buffer);
+    return 0;
+}
+
+static int find_group_indexes_in_cluster_config(SFClusterConfig *cluster,
+        const char *filename)
+{
+    cluster->cluster_group_index = fc_server_get_group_index(
+            &cluster->server_cfg, "cluster");
+    if (cluster->cluster_group_index < 0) {
+        logError("file: "__FILE__", line: %d, "
+                "cluster config file: %s, cluster group "
+                "not configurated", __LINE__, filename);
+        return ENOENT;
+    }
+
+    cluster->service_group_index = fc_server_get_group_index(
+            &cluster->server_cfg, "service");
+    if (cluster->service_group_index < 0) {
+        logError("file: "__FILE__", line: %d, "
+                "cluster config file: %s, service group "
+                "not configurated", __LINE__, filename);
+        return ENOENT;
+    }
+
+    return 0;
+}
+
+static int load_server_cfg(SFClusterConfig *cluster,
+        const char *cluster_filename, const int default_port,
+        const bool share_between_groups)
+{
+    IniContext ini_context;
+    const int min_hosts_each_group = 1;
+    int result;
+
+    if ((result=iniLoadFromFile(cluster_filename, &ini_context)) != 0) {
+        logError("file: "__FILE__", line: %d, "
+                "load conf file \"%s\" fail, ret code: %d",
+                __LINE__, cluster_filename, result);
+        return result;
+    }
+
+    result = fc_server_load_from_ini_context_ex(&cluster->server_cfg,
+            &ini_context, cluster_filename, default_port,
+            min_hosts_each_group, share_between_groups);
+
+    iniFreeContext(&ini_context);
+    return result;
+}
+
+int sf_load_cluster_config_by_file(SFClusterConfig *cluster,
+        const char *full_cluster_filename, const int default_port,
+        const bool share_between_groups, const bool calc_sign)
+{
+    int result;
+
+    if ((result=load_server_cfg(cluster, full_cluster_filename,
+                    default_port, share_between_groups)) != 0)
+    {
+        return result;
+    }
+
+    if ((result=find_group_indexes_in_cluster_config(cluster,
+                    full_cluster_filename)) != 0)
+    {
+        return result;
+    }
+
+    if (calc_sign) {
+        if ((result=calc_cluster_config_sign(cluster)) != 0) {
+            return result;
+        }
+    }
+
+    return 0;
+}
+
+int sf_load_cluster_config_ex1(SFClusterConfig *cluster,
+        IniFullContext *ini_ctx, const char *cluster_config_item_name,
+        const int default_port, char *full_cluster_filename, const int size)
+{
+    const bool share_between_groups = true;
+    char *cluster_config_filename;
+
+    cluster_config_filename = iniGetStrValue(ini_ctx->section_name,
+            cluster_config_item_name, ini_ctx->context);
+    if (cluster_config_filename == NULL || *cluster_config_filename == '\0') {
+        logError("file: "__FILE__", line: %d, "
+                "config file: %s, item \"%s\" not exist or empty",
+                __LINE__, cluster_config_item_name, ini_ctx->filename);
+        return ENOENT;
+    }
+
+    resolve_path(ini_ctx->filename, cluster_config_filename,
+            full_cluster_filename, size);
+    return sf_load_cluster_config_by_file(cluster, full_cluster_filename,
+            default_port, share_between_groups, true);
+}

src/sf_cluster_cfg.h

@@ -0,0 +1,66 @@
+/*
+ * Copyright (c) 2020 YuQing <384681@qq.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <https://www.gnu.org/licenses/>.
+ */
+
+
+#ifndef _SF_CLUSTER_CFG_H
+#define _SF_CLUSTER_CFG_H
+
+#include "sf_types.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+int sf_load_cluster_config_ex1(SFClusterConfig *cluster,
+        IniFullContext *ini_ctx, const char *cluster_config_item_name,
+        const int default_port, char *full_cluster_filename, const int size);
+
+static inline int sf_load_cluster_config_ex(SFClusterConfig *cluster,
+        IniFullContext *ini_ctx, const int default_port,
+        char *full_cluster_filename, const int size)
+{
+    const char *cluster_config_item_name = "cluster_config_filename";
+    return sf_load_cluster_config_ex1(cluster, ini_ctx,
+            cluster_config_item_name, default_port,
+            full_cluster_filename, PATH_MAX);
+}
+
+static inline int sf_load_cluster_config1(SFClusterConfig *cluster,
+        IniFullContext *ini_ctx, const char *cluster_config_item_name,
+        const int default_port)
+{
+    char full_cluster_filename[PATH_MAX];
+    return sf_load_cluster_config_ex1(cluster, ini_ctx,
+            cluster_config_item_name, default_port,
+            full_cluster_filename, PATH_MAX);
+}
+
+static inline int sf_load_cluster_config(SFClusterConfig *cluster,
+        IniFullContext *ini_ctx, const int default_port)
+{
+    char full_cluster_filename[PATH_MAX];
+    return sf_load_cluster_config_ex(cluster, ini_ctx, default_port,
+            full_cluster_filename, PATH_MAX);
+}
+
+int sf_load_cluster_config_by_file(SFClusterConfig *cluster,
+        const char *full_cluster_filename, const int default_port,
+        const bool share_between_groups, const bool calc_sign);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif

src/sf_configs.c

@@ -25,10 +25,10 @@
 #include "fastcommon/logger.h"
 #include "sf_configs.h"
 
-#define DEFAULT_RETRY_MAX_INTERVAL_MS      5000
-#define DEFAULT_CONNECT_RETRY_TIMES          10
+#define DEFAULT_RETRY_MAX_INTERVAL_MS      3000
+#define DEFAULT_CONNECT_RETRY_TIMES         200
 #define DEFAULT_CONNECT_RETRY_INTERVAL_MS   100
-#define DEFAULT_NETWORK_RETRY_TIMES          10
+#define DEFAULT_NETWORK_RETRY_TIMES         200
 #define DEFAULT_NETWORK_RETRY_INTERVAL_MS   100
 
 int sf_load_net_retry_config(SFNetRetryConfig *net_retry_cfg,
@@ -94,13 +94,14 @@ void sf_net_retry_config_to_string(SFNetRetryConfig *net_retry_cfg,
             net_retry_cfg->network.interval_ms);
 }
 
-void sf_load_read_rule_config_ex(SFDataReadRule *rule,
+int sf_load_read_rule_config_ex(SFDataReadRule *rule,
         IniFullContext *ini_ctx, const SFDataReadRule def_rule)
 {
     char *read_rule;
+
     read_rule = iniGetStrValueEx(ini_ctx->section_name,
             "read_rule", ini_ctx->context, true);
-    if (read_rule == NULL || *read_rule == '\0') {
+    if (read_rule == NULL) {
         *rule = def_rule;
     } else if (strncasecmp(read_rule, "any", 3) == 0) {
         *rule = sf_data_read_rule_any_available;
@@ -110,8 +111,62 @@ void sf_load_read_rule_config_ex(SFDataReadRule *rule,
         *rule = sf_data_read_rule_master_only;
     } else {
         logError("file: "__FILE__", line: %d, "
-                "config file: %s, unkown read_rule: %s, set to any",
+                "config file: %s, unkown read_rule: %s",
                 __LINE__, ini_ctx->filename, read_rule);
-        *rule = sf_data_read_rule_any_available;
+        return EINVAL;
     }
+
+    return 0;
+}
+
+int sf_load_election_quorum_config_ex(SFElectionQuorum *quorum,
+        IniFullContext *ini_ctx, const SFElectionQuorum def_quorum)
+{
+    char *str;
+
+    str = iniGetStrValue(ini_ctx->section_name,
+            "quorum", ini_ctx->context);
+    if (str == NULL) {
+        *quorum = def_quorum;
+    } else if (strncasecmp(str, "auto", 4) == 0) {
+        *quorum = sf_election_quorum_auto;
+    } else if (strncasecmp(str, "any", 3) == 0) {
+        *quorum = sf_election_quorum_any;
+    } else if (strncasecmp(str, "majority", 8) == 0) {
+        *quorum = sf_election_quorum_majority;
+    } else {
+        logError("file: "__FILE__", line: %d, "
+                "config file: %s, unkown quorum: %s",
+                __LINE__, ini_ctx->filename, str);
+        return EINVAL;
+    }
+
+    return 0;
+}
+
+int sf_load_replication_quorum_config_ex(SFReplicationQuorum *quorum,
+        IniFullContext *ini_ctx, const SFReplicationQuorum def_quorum)
+{
+    char *str;
+
+    str = iniGetStrValue(ini_ctx->section_name,
+            "quorum", ini_ctx->context);
+    if (str == NULL) {
+        *quorum = def_quorum;
+    } else if (strncasecmp(str, "auto", 4) == 0) {
+        *quorum = sf_replication_quorum_auto;
+    } else if (strncasecmp(str, "any", 3) == 0) {
+        *quorum = sf_replication_quorum_any;
+    } else if (strncasecmp(str, "majority", 8) == 0) {
+        *quorum = sf_replication_quorum_majority;
+    } else if (strncasecmp(str, "smart", 5) == 0) {
+        *quorum = sf_replication_quorum_smart;
+    } else {
+        logError("file: "__FILE__", line: %d, "
+                "config file: %s, unkown quorum: %s",
+                __LINE__, ini_ctx->filename, str);
+        return EINVAL;
+    }
+
+    return 0;
 }

src/sf_configs.h

@@ -23,39 +23,12 @@
 #include "sf_define.h"
 #include "sf_types.h"
 
-typedef enum sf_net_retry_interval_mode {
-    sf_net_retry_interval_mode_fixed,
-    sf_net_retry_interval_mode_multiple
-} SFNetRetryIntervalMode;
-
-typedef struct sf_net_retry_interval_mode_max_pair {
-    SFNetRetryIntervalMode mode;
-    int max_interval_ms;
-} SFNetRetryIntervalModeMaxPair;
-
-typedef struct sf_net_retry_times_interval_pair {
-    int times;
-    int interval_ms;
-} SFNetRetryTimesIntervalPair;
-
-typedef struct sf_net_retry_config {
-    SFNetRetryIntervalModeMaxPair interval_mm;
-    SFNetRetryTimesIntervalPair connect;
-    SFNetRetryTimesIntervalPair network;
-} SFNetRetryConfig;
-
 typedef struct sf_net_retry_interval_context {
-    SFNetRetryIntervalModeMaxPair *mm;
-    SFNetRetryTimesIntervalPair *ti;
+    const SFNetRetryIntervalModeMaxPair *mm;
+    const SFNetRetryTimesIntervalPair *ti;
     int interval_ms;
 } SFNetRetryIntervalContext;
 
-typedef enum sf_data_read_rule {
-    sf_data_read_rule_any_available,
-    sf_data_read_rule_slave_first,
-    sf_data_read_rule_master_only,
-} SFDataReadRule;
-
 #ifdef __cplusplus
 extern "C" {
 #endif
@@ -72,8 +45,8 @@ static inline void sf_reset_net_retry_interval(SFNetRetryIntervalContext *ctx)
 }
 
 static inline void sf_init_net_retry_interval_context(
-        SFNetRetryIntervalContext *ctx, SFNetRetryIntervalModeMaxPair *mm,
-        SFNetRetryTimesIntervalPair *ti)
+        SFNetRetryIntervalContext *ctx, const SFNetRetryIntervalModeMaxPair *mm,
+        const SFNetRetryTimesIntervalPair *ti)
 {
     ctx->mm = mm;
     ctx->ti = ti;
@@ -94,7 +67,7 @@ static inline int sf_calc_next_retry_interval(SFNetRetryIntervalContext *ctx)
     return ctx->interval_ms;
 }
 
-void sf_load_read_rule_config_ex(SFDataReadRule *rule,
+int sf_load_read_rule_config_ex(SFDataReadRule *rule,
         IniFullContext *ini_ctx, const SFDataReadRule def_rule);
 
 static inline const char *sf_get_read_rule_caption(
@@ -112,9 +85,113 @@ static inline const char *sf_get_read_rule_caption(
     }
 }
 
+int sf_load_election_quorum_config_ex(SFElectionQuorum *quorum,
+        IniFullContext *ini_ctx, const SFElectionQuorum def_quorum);
+
+static inline const char *sf_get_election_quorum_caption(
+        const SFElectionQuorum quorum)
+{
+    switch (quorum) {
+        case sf_election_quorum_auto:
+            return "auto";
+        case sf_election_quorum_any:
+            return "any";
+        case sf_election_quorum_majority:
+            return "majority";
+        default:
+            return "unknown";
+    }
+}
+
+static inline bool sf_election_quorum_check(const SFElectionQuorum quorum,
+        const bool vote_node_enabled, const int total_count,
+        const int active_count)
+{
+    switch (quorum) {
+        case sf_election_quorum_any:
+            return active_count > 0;
+        case sf_election_quorum_auto:
+            if (total_count % 2 == 0 && !vote_node_enabled) {
+                return active_count > 0;  //same as sf_election_quorum_any
+            }
+            //continue
+        case sf_election_quorum_majority:
+            if (active_count == total_count) {
+                return true;
+            } else {
+                return active_count > total_count / 2;
+            }
+    }
+}
+
+int sf_load_replication_quorum_config_ex(SFReplicationQuorum *quorum,
+        IniFullContext *ini_ctx, const SFReplicationQuorum def_quorum);
+
+static inline const char *sf_get_replication_quorum_caption(
+        const SFReplicationQuorum quorum)
+{
+    switch (quorum) {
+        case sf_replication_quorum_auto:
+            return "auto";
+        case sf_replication_quorum_any:
+            return "any";
+        case sf_replication_quorum_majority:
+            return "majority";
+        case sf_replication_quorum_smart:
+            return "smart";
+        default:
+            return "unknown";
+    }
+}
+
+#define SF_REPLICATION_QUORUM_MAJORITY(server_count, success_count) \
+    ((success_count == server_count) || (success_count > server_count / 2))
+
+static inline bool sf_replication_quorum_check(const SFReplicationQuorum
+        quorum, const int server_count, const int success_count)
+{
+    switch (quorum) {
+        case sf_replication_quorum_any:
+            return true;
+        case sf_replication_quorum_auto:
+            if (server_count % 2 == 0) {
+                return true;  //same as sf_replication_quorum_any
+            }
+            //continue
+        case sf_replication_quorum_smart:
+        case sf_replication_quorum_majority:
+            return SF_REPLICATION_QUORUM_MAJORITY(
+                    server_count, success_count);
+    }
+}
+
 #define sf_load_read_rule_config(rule, ini_ctx) \
     sf_load_read_rule_config_ex(rule, ini_ctx, sf_data_read_rule_master_only)
 
+#define sf_load_election_quorum_config(quorum, ini_ctx) \
+    sf_load_election_quorum_config_ex(quorum, ini_ctx, sf_election_quorum_auto)
+
+#define sf_load_replication_quorum_config(quorum, ini_ctx) \
+    sf_load_replication_quorum_config_ex(quorum, ini_ctx,  \
+            sf_replication_quorum_auto)
+
+#define SF_ELECTION_QUORUM_NEED_REQUEST_VOTE_NODE(quorum, \
+        vote_node_enabled, server_count, active_count) \
+    (active_count < server_count && vote_node_enabled && \
+     quorum != sf_election_quorum_any && server_count % 2 == 0)
+
+#define SF_ELECTION_QUORUM_NEED_CHECK_VOTE_NODE(quorum, \
+        vote_node_enabled, server_count) \
+    (vote_node_enabled && quorum != sf_election_quorum_any \
+     && server_count % 2 == 0)
+
+#define SF_REPLICATION_QUORUM_NEED_MAJORITY(quorum, server_count) \
+    (server_count > 1 && (quorum != sf_replication_quorum_any))
+
+#define SF_REPLICATION_QUORUM_NEED_DETECT(quorum, server_count)  \
+    (server_count % 2 == 0 && (quorum == sf_replication_quorum_smart || \
+                               quorum == sf_replication_quorum_auto))
+
 #define SF_NET_RETRY_FINISHED(retry_times, counter, result)  \
         !((SF_IS_RETRIABLE_ERROR(result) && ((retry_times > 0 &&  \
                     counter <= retry_times) || (retry_times < 0))))

src/sf_connection_manager.c

@@ -0,0 +1,971 @@
+/*
+ * Copyright (c) 2020 YuQing <384681@qq.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <https://www.gnu.org/licenses/>.
+ */
+
+#include <time.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <ctype.h>
+#include <unistd.h>
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <fcntl.h>
+#include <errno.h>
+#include "sf_global.h"
+#include "sf_configs.h"
+#include "sf_proto.h"
+#include "fastcommon/shared_func.h"
+#include "fastcommon/logger.h"
+#include "fastcommon/fc_atomic.h"
+#include "sf_connection_manager.h"
+
+static int get_group_servers(SFConnectionManager *cm,
+        SFCMConnGroupEntry *group);
+
+static ConnectionInfo *get_spec_connection(SFConnectionManager *cm,
+        const ConnectionInfo *target, const bool shared, int *err_no)
+{
+    return conn_pool_get_connection_ex(&cm->cpool,
+            target, cm->module_name, shared, err_no);
+}
+
+static ConnectionInfo *make_connection(SFConnectionManager *cm,
+        FCAddressPtrArray *addr_array, const bool shared, int *err_no)
+{
+    FCAddressInfo **current;
+    FCAddressInfo **addr;
+    FCAddressInfo **end;
+    ConnectionInfo *conn;
+
+    if (addr_array->count <= 0) {
+        *err_no = ENOENT;
+        return NULL;
+    }
+
+    current = addr_array->addrs + addr_array->index;
+    if ((conn=get_spec_connection(cm, &(*current)->conn,
+                    shared, err_no)) != NULL)
+    {
+        return conn;
+    }
+
+    if (addr_array->count == 1) {
+        return NULL;
+    }
+
+    end = addr_array->addrs + addr_array->count;
+    for (addr=addr_array->addrs; addr<end; addr++) {
+        if (addr == current) {
+            continue;
+        }
+
+        if ((conn=get_spec_connection(cm, &(*addr)->conn,
+                        shared, err_no)) != NULL)
+        {
+            addr_array->index = addr - addr_array->addrs;
+            return conn;
+        }
+    }
+
+    return NULL;
+}
+
+static ConnectionInfo *get_server_connection(SFConnectionManager *cm,
+        FCServerInfo *server, const bool shared, int *err_no)
+{
+    FCAddressPtrArray *addr_array;
+    ConnectionInfo *conn;
+
+    addr_array = &server->group_addrs[cm->server_group_index].address_array;
+    if ((conn=make_connection(cm, addr_array, shared, err_no)) == NULL) {
+        logError("file: "__FILE__", line: %d, "
+                "%s server id: %d, address count: %d, get_server_connection fail",
+                __LINE__, cm->module_name, server->id, addr_array->count);
+    }
+    return conn;
+}
+
+static ConnectionInfo *get_connection(SFConnectionManager *cm,
+        const int group_index, const bool shared, int *err_no)
+{
+    SFCMServerArray *server_array;
+    ConnectionInfo *conn;
+    uint32_t server_hash_code;
+    int server_index;
+    int i;
+
+    server_array = &cm->groups.entries[group_index].all;
+    server_hash_code = rand();
+    server_index = server_hash_code % server_array->count;
+    if ((conn=make_connection(cm, server_array->servers[server_index].
+                    addr_array, shared, err_no)) != NULL)
+    {
+        return conn;
+    }
+
+    if (server_array->count > 1) {
+        for (i=0; i<server_array->count; i++) {
+            if (i == server_index) {
+                continue;
+            }
+
+            if ((conn=make_connection(cm, server_array->servers[i].
+                            addr_array, shared, err_no)) != NULL)
+            {
+                return conn;
+            }
+        }
+    }
+
+    logError("file: "__FILE__", line: %d, "
+            "%s data group index: %d, get_connection fail, "
+            "configured server count: %d", __LINE__, cm->module_name,
+            group_index, server_array->count);
+    return NULL;
+}
+
+static inline void set_connection_params(ConnectionInfo *conn,
+        SFCMServerEntry *server, SFCMServerPtrArray *old_alives)
+{
+    SFConnectionParameters *cparam;
+    cparam = (SFConnectionParameters *)conn->args;
+    cparam->cm.sentry = server;
+    cparam->cm.old_alives = old_alives;
+}
+
+static inline int push_to_detect_queue(SFConnectionManager *cm,
+        SFCMConnGroupEntry *group, SFCMServerPtrArray *alives)
+{
+    if (!cm->alive_detect.bg_thread_enabled) {
+        return 0;
+    }
+
+    if (alives->count < group->all.count) {
+        if (__sync_bool_compare_and_swap(&group->in_queue, 0, 1)) {
+            return common_blocked_queue_push(&cm->alive_detect.queue, group);
+        }
+    }
+
+    return 0;
+}
+
+static inline bool alive_array_cas(SFConnectionManager *cm,
+        SFCMConnGroupEntry *group, SFCMServerPtrArray *old_alives,
+        SFCMServerPtrArray *new_alives)
+{
+    if (__sync_bool_compare_and_swap(&group->alives,
+                old_alives, new_alives))
+    {
+        logDebug("file: "__FILE__", line: %d, "
+                "[%s] group_id: %d, old alive server count: %d, "
+                "new alive server count: %d", __LINE__, cm->module_name,
+                group->id, old_alives->count, new_alives->count);
+
+        push_to_detect_queue(cm, group, new_alives);
+        fast_mblock_delay_free_object(&cm->sptr_array_allocator, old_alives,
+                (cm->common_cfg->connect_timeout + cm->common_cfg->
+                 network_timeout) * group->all.count * 4);
+        return true;
+    } else {
+        fast_mblock_free_object(&cm->sptr_array_allocator, new_alives);
+        return false;
+    }
+}
+
+static int remove_from_alives(SFConnectionManager *cm,
+        SFCMConnGroupEntry *group, SFCMServerPtrArray *old_alives,
+        SFCMServerEntry *server)
+{
+    SFCMServerPtrArray *new_alives;
+    SFCMServerEntry **pp;
+    SFCMServerEntry **dest;
+    SFCMServerEntry **end;
+
+    new_alives = (SFCMServerPtrArray *)FC_ATOMIC_GET(group->alives);
+    if (new_alives != old_alives) {
+        return 0;
+    }
+
+    new_alives = (SFCMServerPtrArray *)fast_mblock_alloc_object(
+            &cm->sptr_array_allocator);
+    if (new_alives == NULL) {
+        return ENOMEM;
+    }
+
+    dest = new_alives->servers;
+    end = old_alives->servers + old_alives->count;
+    for (pp=old_alives->servers; pp<end; pp++) {
+        if (*pp != server) {
+            *dest++ = *pp;
+        }
+    }
+
+    new_alives->count = dest - new_alives->servers;
+    if (alive_array_cas(cm, group, old_alives, new_alives)) {
+        SFCMServerEntry *master;
+        master = (SFCMServerEntry *)FC_ATOMIC_GET(group->master);
+        if (master == server) {
+            __sync_bool_compare_and_swap(&group->master, master, NULL);
+        }
+    }
+
+    return 0;
+}
+
+static inline ConnectionInfo *make_master_connection(SFConnectionManager *cm,
+        SFCMConnGroupEntry *group, const bool shared, int *err_no)
+{
+    SFCMServerEntry *master;
+    ConnectionInfo *conn;
+    SFCMServerPtrArray *alives;
+
+    master = (SFCMServerEntry *)FC_ATOMIC_GET(group->master);
+    if (master != NULL) {
+        if ((conn=make_connection(cm, master->addr_array,
+                        shared, err_no)) != NULL)
+        {
+            alives = (SFCMServerPtrArray *)FC_ATOMIC_GET(group->alives);
+            set_connection_params(conn, master, alives);
+            return conn;
+        } else {
+            alives = (SFCMServerPtrArray *)FC_ATOMIC_GET(group->alives);
+            if (alives != NULL) {
+                remove_from_alives(cm, group, alives, master);
+            }
+            __sync_bool_compare_and_swap(&group->master, master, NULL);
+        }
+    }
+
+    *err_no = SF_RETRIABLE_ERROR_NO_SERVER;
+    return NULL;
+}
+
+static inline ConnectionInfo *make_readable_connection(SFConnectionManager *cm,
+        SFCMConnGroupEntry *group, SFCMServerPtrArray *alives,
+        const int index, const bool shared, int *err_no)
+{
+    ConnectionInfo *conn;
+
+    if ((conn=make_connection(cm, alives->servers[index]->
+                    addr_array, shared, err_no)) == NULL)
+    {
+        remove_from_alives(cm, group, alives, alives->servers[index]);
+    } else {
+        set_connection_params(conn, alives->servers[index], alives);
+    }
+
+    return conn;
+}
+
+static ConnectionInfo *get_master_connection(SFConnectionManager *cm,
+        const int group_index, const bool shared, int *err_no)
+{
+    SFCMConnGroupEntry *group;
+    ConnectionInfo *conn;
+    SFNetRetryIntervalContext net_retry_ctx;
+    int retry_count;
+
+    group = cm->groups.entries + group_index;
+    sf_init_net_retry_interval_context(&net_retry_ctx,
+            &cm->common_cfg->net_retry_cfg.interval_mm,
+            &cm->common_cfg->net_retry_cfg.connect);
+    retry_count = 0;
+    while (1) {
+        if ((conn=make_master_connection(cm, group, shared, err_no)) != NULL) {
+            return conn;
+        }
+
+        /*
+        logInfo("file: "__FILE__", line: %d, "
+                "retry_count: %d, interval_ms: %d, data group id: %d, "
+                "master: %p, alive count: %d, all count: %d", __LINE__,
+                retry_count, net_retry_ctx.interval_ms, group->id,
+                FC_ATOMIC_GET(group->master), ((SFCMServerPtrArray *)
+                    FC_ATOMIC_GET(group->alives))->count, group->all.count);
+         */
+
+        *err_no = get_group_servers(cm, group);
+        if (*err_no == 0) {
+            *err_no = SF_RETRIABLE_ERROR_NO_SERVER;  //for try again
+        }
+        SF_NET_RETRY_CHECK_AND_SLEEP(net_retry_ctx,
+                cm->common_cfg->net_retry_cfg.
+                connect.times, ++retry_count, *err_no);
+    }
+
+    logError("file: "__FILE__", line: %d, "
+            "%s get_master_connection fail, group id: %d, "
+            "retry count: %d, " "errno: %d", __LINE__,
+            cm->module_name, group->id, retry_count, *err_no);
+    return NULL;
+}
+
+static ConnectionInfo *get_readable_connection(SFConnectionManager *cm,
+        const int group_index, const bool shared, int *err_no)
+{
+    SFCMConnGroupEntry *group;
+    SFCMServerPtrArray *alives;
+    ConnectionInfo *conn;
+    SFNetRetryIntervalContext net_retry_ctx;
+    uint32_t index;
+    int retry_count;
+
+    group = cm->groups.entries + group_index;
+    if ((cm->common_cfg->read_rule == sf_data_read_rule_master_only) ||
+            (group->all.count == 1))
+    {
+        return get_master_connection(cm, group_index, shared, err_no);
+    }
+
+    sf_init_net_retry_interval_context(&net_retry_ctx,
+            &cm->common_cfg->net_retry_cfg.interval_mm,
+            &cm->common_cfg->net_retry_cfg.connect);
+    retry_count = 0;
+    while (1) {
+        alives = (SFCMServerPtrArray *)FC_ATOMIC_GET(group->alives);
+        if (alives->count > 0) {
+            index = rand() % alives->count;
+            if ((conn=make_readable_connection(cm, group, alives,
+                            index, shared, err_no)) != NULL)
+            {
+                return conn;
+            }
+        }
+
+        if (cm->common_cfg->read_rule == sf_data_read_rule_slave_first) {
+            if ((conn=make_master_connection(cm, group, shared, err_no)) != NULL) {
+                return conn;
+            }
+        }
+
+        *err_no = get_group_servers(cm, group);
+        if (*err_no == 0) {
+            *err_no = SF_RETRIABLE_ERROR_NO_SERVER;  //for try again
+        }
+        SF_NET_RETRY_CHECK_AND_SLEEP(net_retry_ctx,
+                cm->common_cfg->net_retry_cfg.
+                connect.times, ++retry_count, *err_no);
+    }
+
+    logError("file: "__FILE__", line: %d, "
+            "%s get_readable_connection fail, retry count: %d, errno: %d",
+            __LINE__, cm->module_name, retry_count, *err_no);
+    return NULL;
+}
+
+static void release_connection(SFConnectionManager *cm,
+        ConnectionInfo *conn)
+{
+    SFConnectionParameters *cparam;
+    cparam = (SFConnectionParameters *)conn->args;
+    if (cparam->cm.sentry != NULL) {
+        cparam->cm.sentry = NULL;
+        cparam->cm.old_alives = NULL;
+    }
+
+    conn_pool_close_connection_ex(&cm->cpool, conn, false);
+}
+
+static void close_connection(SFConnectionManager *cm, ConnectionInfo *conn)
+{
+    SFConnectionParameters *cparam;
+    SFCMServerEntry *server;
+    SFCMConnGroupEntry *group;
+
+    cparam = (SFConnectionParameters *)conn->args;
+    if (cparam->cm.sentry != NULL) {
+        server = cparam->cm.sentry;
+        group = cm->groups.entries + server->group_index;
+        if (cparam->cm.old_alives != NULL) {
+            remove_from_alives(cm, group, cparam->cm.old_alives, server);
+            cparam->cm.old_alives = NULL;
+        }
+        __sync_bool_compare_and_swap(&group->master, server, NULL);
+        cparam->cm.sentry = NULL;
+    }
+
+    conn_pool_close_connection_ex(&cm->cpool, conn, true);
+}
+
+static ConnectionInfo *get_leader_connection(SFConnectionManager *cm,
+        FCServerInfo *server, const bool shared, int *err_no)
+{
+    ConnectionInfo *conn;
+    SFClientServerEntry leader;
+    SFNetRetryIntervalContext net_retry_ctx;
+    char formatted_ip[FORMATTED_IP_SIZE];
+    int i;
+    int connect_fails;
+
+    sf_init_net_retry_interval_context(&net_retry_ctx,
+            &cm->common_cfg->net_retry_cfg.interval_mm,
+            &cm->common_cfg->net_retry_cfg.connect);
+    i = connect_fails = 0;
+    while (1) {
+        do {
+            if ((conn=get_server_connection(cm, server,
+                            shared, err_no)) == NULL)
+            {
+                connect_fails++;
+                break;
+            }
+
+            if ((*err_no=sf_proto_get_leader(conn, cm->module_name, cm->
+                            common_cfg->network_timeout, &leader)) != 0)
+            {
+                close_connection(cm, conn);
+                break;
+            }
+
+            if (FC_CONNECTION_SERVER_EQUAL1(*conn, leader.conn)) {
+                return conn;
+            }
+            release_connection(cm, conn);
+
+            if ((conn=get_spec_connection(cm, &leader.conn,
+                            shared, err_no)) == NULL)
+            {
+                if (cm->server_cfg != NULL) {
+                    FCServerInfo *ls;
+                    if ((ls=fc_server_get_by_id(cm->server_cfg,
+                                    leader.server_id)) != NULL)
+                    {
+                        if (ls->group_addrs[cm->server_group_index].
+                                address_array.count > 1)
+                        {
+                            if ((conn=get_server_connection(cm, ls,
+                                            shared, err_no)) != NULL)
+                            {
+                                return conn;
+                            }
+                        }
+                    }
+                }
+
+                break;
+            }
+
+            return conn;
+        } while (0);
+
+        if (connect_fails == 2) {
+            break;
+        }
+
+        SF_NET_RETRY_CHECK_AND_SLEEP(net_retry_ctx,
+                cm->common_cfg->net_retry_cfg.
+                connect.times, ++i, *err_no);
+    }
+
+    format_ip_address(server->group_addrs[cm->server_group_index].
+            address_array.addrs[0]->conn.ip_addr, formatted_ip);
+    logWarning("file: "__FILE__", line: %d, "
+            "%s get_leader_connection fail, server id: %d, %s:%u, errno: %d",
+            __LINE__, cm->module_name, server->id, formatted_ip,
+            server->group_addrs[cm->server_group_index].address_array.
+            addrs[0]->conn.port, *err_no);
+    return NULL;
+}
+
+const struct sf_connection_parameters *sf_cm_get_connection_params(
+        SFConnectionManager *cm, ConnectionInfo *conn)
+{
+    return (SFConnectionParameters *)conn->args;
+}
+
+int sf_cm_validate_connection_callback(ConnectionInfo *conn, void *args)
+{
+    SFConnectionManager *cm;
+    SFResponseInfo response;
+    int result;
+
+    cm = (SFConnectionManager *)args;
+    if ((result=sf_active_test(conn, &response, cm->common_cfg->
+                    network_timeout)) != 0)
+    {
+        sf_log_network_error(&response, conn, cm->module_name, result);
+    }
+
+    return result;
+}
+
+static int init_group_array(SFConnectionManager *cm,
+        SFCMConnGroupArray *garray, const int group_count)
+{
+    int bytes;
+
+    bytes = sizeof(SFCMConnGroupEntry) * group_count;
+    garray->entries = (SFCMConnGroupEntry *)fc_malloc(bytes);
+    if (garray->entries == NULL) {
+        return ENOMEM;
+    }
+    memset(garray->entries, 0, bytes);
+    garray->count = group_count;
+    return 0;
+}
+
+int sf_connection_manager_init_ex(SFConnectionManager *cm,
+        const char *module_name, const SFClientCommonConfig *common_cfg,
+        const int group_count, const int server_group_index,
+        const int server_count, const int max_count_per_entry,
+        const int max_idle_time, fc_connection_callback_func
+        connect_done_callback, void *args, FCServerConfig *server_cfg,
+        const bool bg_thread_enabled)
+{
+    struct {
+        ConnectionExtraParams holder;
+        ConnectionExtraParams *ptr;
+    } extra_params;
+    FCServerGroupInfo *server_group;
+
+    int htable_capacity;
+    int result;
+
+    if (server_count <= 4) {
+        htable_capacity = 16;
+    } else if (server_count <= 16) {
+        htable_capacity = 64;
+    } else if (server_count <= 32) {
+        htable_capacity = 128;
+    } else if (server_count < 64) {
+        htable_capacity = 256;
+    } else {
+        htable_capacity = 4 * server_count;
+    }
+
+    if ((server_group=fc_server_get_group_by_index(server_cfg,
+                    server_group_index)) == NULL)
+    {
+        return ENOENT;
+    }
+
+    if (server_group->comm_type == fc_comm_type_sock) {
+        extra_params.ptr = NULL;
+    } else {
+        if ((result=conn_pool_set_rdma_extra_params(&extra_params.holder,
+                        server_cfg, server_group_index)) != 0)
+        {
+            return result;
+        }
+        extra_params.ptr = &extra_params.holder;
+    }
+    if ((result=conn_pool_init_ex1(&cm->cpool, common_cfg->connect_timeout,
+                    max_count_per_entry, max_idle_time, htable_capacity,
+                    connect_done_callback, args,
+                    sf_cm_validate_connection_callback, cm,
+                    sizeof(SFConnectionParameters),
+                    extra_params.ptr)) != 0)
+    {
+        return result;
+    }
+
+    if ((result=init_group_array(cm, &cm->groups, group_count)) != 0) {
+        return result;
+    }
+
+    if (bg_thread_enabled) {
+        if ((result=common_blocked_queue_init(&cm->
+                        alive_detect.queue)) != 0)
+        {
+            return result;
+        }
+    }
+
+    cm->server_group_index = server_group_index;
+    cm->module_name = module_name;
+    cm->common_cfg = common_cfg;
+    cm->server_cfg = server_cfg;
+    cm->alive_detect.bg_thread_enabled = bg_thread_enabled;
+    cm->max_servers_per_group = 0;
+    cm->extra = NULL;
+    cm->exclude_server_id = 0;
+
+    cm->ops.get_connection = get_connection;
+    cm->ops.get_server_connection = get_server_connection;
+    cm->ops.get_spec_connection = get_spec_connection;
+    cm->ops.get_master_connection = get_master_connection;
+    cm->ops.get_readable_connection = get_readable_connection;
+    cm->ops.get_leader_connection = get_leader_connection;
+    cm->ops.release_connection = release_connection;
+    cm->ops.close_connection = close_connection;
+    cm->ops.get_connection_params = sf_cm_get_connection_params;
+    return 0;
+}
+
+int sf_connection_manager_add(SFConnectionManager *cm, const int group_id,
+        FCServerInfo **servers, const int count)
+{
+    SFCMConnGroupEntry *group;
+    FCServerInfo **server;
+    FCServerInfo **end;
+    SFCMServerEntry *entry;
+    int group_index;
+
+    if (group_id < 1) {
+        logError("file: "__FILE__", line: %d, "
+                "invalid group id: %d < 1",
+                __LINE__, group_id);
+        return EINVAL;
+    }
+    if (group_id > cm->groups.count) {
+        logError("file: "__FILE__", line: %d, "
+                "invalid group id: %d > group count: %d",
+                __LINE__, group_id, cm->groups.count);
+        return EINVAL;
+    }
+
+    group_index = group_id - 1;
+    group = cm->groups.entries + group_index;
+    group->id = group_id;
+    group->all.servers = (SFCMServerEntry *)fc_malloc(
+            sizeof(SFCMServerEntry) * count);
+    if (group->all.servers == NULL) {
+        return ENOMEM;
+    }
+    group->all.count = count;
+
+    end = servers + count;
+    for (entry=group->all.servers, server=servers;
+            server<end; entry++, server++)
+    {
+        entry->id = (*server)->id;
+        entry->group_index = group_index;
+        entry->addr_array = &(*server)->group_addrs[
+            cm->server_group_index].address_array;
+    }
+
+    if (count > cm->max_servers_per_group) {
+        cm->max_servers_per_group = count;
+    }
+
+    return 0;
+}
+
+static SFCMServerEntry *get_server_by_id(SFCMConnGroupEntry *group,
+        const int server_id)
+{
+    SFCMServerEntry *server;
+    SFCMServerEntry *end;
+
+    end = group->all.servers + group->all.count;
+    for (server=group->all.servers; server<end; server++) {
+        if (server->id == server_id) {
+            return server;
+        }
+    }
+
+    return NULL;
+}
+
+static SFCMServerPtrArray *convert_to_sptr_array(SFConnectionManager *cm,
+        SFCMConnGroupEntry *group, SFGroupServerArray *sarray, int *err_no)
+{
+    SFCMServerPtrArray *alives;
+    SFGroupServerInfo *server;
+    SFGroupServerInfo *end;
+    SFCMServerEntry *sentry;
+
+    if (sarray->count > cm->max_servers_per_group) {
+        logError("file: "__FILE__", line: %d, "
+                "group id: %d, response server count: %d > "
+                "max count: %d!", __LINE__, group->id,
+                sarray->count, cm->max_servers_per_group);
+        *err_no = EOVERFLOW;
+        return NULL;
+    }
+
+    alives = (SFCMServerPtrArray *)fast_mblock_alloc_object(
+            &cm->sptr_array_allocator);
+    if (alives == NULL) {
+        *err_no = ENOMEM;
+        return NULL;
+    }
+
+    alives->count = 0;
+    end = sarray->servers + sarray->count;
+    for (server=sarray->servers; server<end; server++) {
+        if ((sentry=get_server_by_id(group, server->id)) == NULL) {
+            logError("file: "__FILE__", line: %d, "
+                    "group id: %d, response server count: %d > "
+                    "max count: %d!", __LINE__, group->id,
+                    sarray->count, cm->max_servers_per_group);
+            *err_no = ENOENT;
+            fast_mblock_free_object(&cm->sptr_array_allocator, alives);
+            return NULL;
+        }
+
+        if (server->is_master) {
+            FC_ATOMIC_SET(group->master, sentry);
+            if (cm->common_cfg->read_rule != sf_data_read_rule_slave_first) {
+                alives->servers[alives->count++] = sentry;
+            }
+        } else if (server->is_active) {
+            alives->servers[alives->count++] = sentry;
+        }
+    }
+
+    *err_no = 0;
+    return alives;
+}
+
+static int sptr_array_compare(SFCMServerPtrArray *a1,
+        SFCMServerPtrArray *a2)
+{
+    int sub;
+    int i;
+
+    if ((sub=(a1->count - a2->count)) != 0) {
+        return sub;
+    }
+
+    for (i = 0; i < a1->count; i++) {
+        if ((sub=(a1->servers[i]->id - a2->servers[i]->id)) != 0) {
+            return sub;
+        }
+    }
+
+    return 0;
+}
+
+static int do_get_group_servers(SFConnectionManager *cm,
+        SFCMConnGroupEntry *group, ConnectionInfo *conn)
+{
+#define MAX_GROUP_SERVER_COUNT 128
+    int result;
+    SFGroupServerInfo fixed_servers[MAX_GROUP_SERVER_COUNT];
+    SFGroupServerArray sarray;
+    SFCMServerPtrArray *old_alives;
+    SFCMServerPtrArray *new_alives;
+
+    sarray.alloc = MAX_GROUP_SERVER_COUNT;
+    sarray.count = 0;
+    sarray.servers = fixed_servers;
+    if ((result=sf_proto_get_group_servers(conn, cm->module_name, cm->
+                    common_cfg->network_timeout, group->id, &sarray)) != 0)
+    {
+        return result;
+    }
+
+    if ((new_alives=convert_to_sptr_array(cm, group,
+                    &sarray, &result)) == NULL)
+    {
+        return result;
+    }
+
+    old_alives = (SFCMServerPtrArray *)FC_ATOMIC_GET(group->alives);
+    if (sptr_array_compare(old_alives, new_alives) == 0) {
+        push_to_detect_queue(cm, group, new_alives);
+        fast_mblock_free_object(&cm->sptr_array_allocator, new_alives);
+        return 0;
+    }
+
+    alive_array_cas(cm, group, old_alives, new_alives);
+    return 0;
+}
+
+static int get_group_servers_by_active(SFConnectionManager *cm,
+        SFCMConnGroupEntry *group)
+{
+    const bool shared = true;
+    SFCMServerPtrArray *alives;
+    SFCMServerEntry **server;
+    SFCMServerEntry **end;
+    ConnectionInfo *conn;
+    int result;
+
+    result = ENOENT;
+    alives = (SFCMServerPtrArray *)FC_ATOMIC_GET(group->alives);
+    if (alives->count == 0) {
+        return result;
+    }
+
+    end = alives->servers + alives->count;
+    for (server=alives->servers; server<end; server++) {
+        if ((conn=make_connection(cm, (*server)->addr_array,
+                        shared, &result)) == NULL)
+        {
+            continue;
+        }
+
+        result = do_get_group_servers(cm, group, conn);
+        conn_pool_close_connection_ex(&cm->cpool, conn, result != 0);
+        if (result == 0) {
+            return 0;
+        }
+    }
+
+    return result;
+}
+
+static int get_group_servers_by_all(SFConnectionManager *cm,
+        SFCMConnGroupEntry *group)
+{
+    const bool shared = true;
+    SFCMServerEntry *server;
+    SFCMServerEntry *end;
+    ConnectionInfo *conn;
+    int result;
+
+    result = ENOENT;
+    if (group->all.count == 0) {
+        return result;
+    }
+
+    end = group->all.servers + group->all.count;
+    for (server=group->all.servers; server<end; server++) {
+        if (server->id == cm->exclude_server_id) {
+            continue;
+        }
+
+        if ((conn=make_connection(cm, server->addr_array,
+                        shared, &result)) == NULL)
+        {
+            continue;
+        }
+
+        result = do_get_group_servers(cm, group, conn);
+        conn_pool_close_connection_ex(&cm->cpool, conn, result != 0);
+        if (result == 0) {
+            return 0;
+        }
+    }
+
+    return result;
+}
+
+static int get_group_servers(SFConnectionManager *cm,
+        SFCMConnGroupEntry *group)
+{
+    int result;
+
+    if ((result=get_group_servers_by_active(cm, group)) == 0) {
+        return 0;
+    }
+
+    return get_group_servers_by_all(cm, group);
+}
+
+static void deal_nodes(SFConnectionManager *cm,
+        struct common_blocked_node *node)
+{
+    SFCMConnGroupEntry *group;
+    SFCMServerPtrArray *alives;
+
+    do {
+        group = (SFCMConnGroupEntry *)node->data;
+        __sync_bool_compare_and_swap(&group->in_queue, 1, 0);
+        alives = (SFCMServerPtrArray *)FC_ATOMIC_GET(group->alives);
+        if (alives->count < group->all.count) {
+            logDebug("file: "__FILE__", line: %d, "
+                    "[%s] group_id: %d, alive server count: %d, "
+                    "all server count: %d", __LINE__, cm->module_name,
+                    group->id, alives->count, group->all.count);
+
+            if (get_group_servers(cm, group) != 0) {
+                push_to_detect_queue(cm, group, (SFCMServerPtrArray *)
+                        FC_ATOMIC_GET(group->alives));
+            }
+        }
+
+        node = node->next;
+    } while (node != NULL);
+}
+
+static void *connection_manager_thread_func(void *arg)
+{
+    SFConnectionManager *cm;
+    struct common_blocked_node *head;
+
+#ifdef OS_LINUX
+    prctl(PR_SET_NAME, "cm-alive-detect");
+#endif
+
+    cm = (SFConnectionManager *)arg;
+    logDebug("file: "__FILE__", line: %d, "
+            "[%s] connection manager thread start",
+            __LINE__, cm->module_name);
+
+    while (1) {
+        sleep(1);
+        if ((head=common_blocked_queue_pop_all_nodes(&cm->
+                        alive_detect.queue)) == NULL)
+        {
+            continue;
+        }
+
+        deal_nodes(cm, head);
+        common_blocked_queue_free_all_nodes(&cm->alive_detect.queue, head);
+    }
+
+    return NULL;
+}
+
+static int sptr_array_alloc_init(void *element, void *args)
+{
+    SFCMServerPtrArray *sptr_array;
+
+    sptr_array = (SFCMServerPtrArray *)element;
+    sptr_array->servers = (SFCMServerEntry **)(sptr_array + 1);
+    return 0;
+}
+
+int sf_connection_manager_prepare(SFConnectionManager *cm)
+{
+    int result;
+    int element_size;
+    SFCMConnGroupEntry *group;
+    SFCMConnGroupEntry *end;
+    SFCMServerPtrArray *sptr_array;
+
+    element_size = sizeof(SFCMServerPtrArray) +
+        sizeof(SFCMServerEntry *) * cm->max_servers_per_group;
+    if ((result=fast_mblock_init_ex1(&cm->sptr_array_allocator,
+                    "server-ptr-array", element_size, 4 * 1024, 0,
+                    sptr_array_alloc_init, NULL, true)) != 0)
+    {
+        return result;
+    }
+
+    end = cm->groups.entries + cm->groups.count;
+    for (group=cm->groups.entries; group<end; group++) {
+        if (group->all.count == 0) {
+            logError("file: "__FILE__", line: %d, "
+                    "group id: %d, no servers!",
+                    __LINE__, group->id);
+            return ENOENT;
+        }
+
+        sptr_array = (SFCMServerPtrArray *)fast_mblock_alloc_object(
+                &cm->sptr_array_allocator);
+        if (sptr_array == NULL) {
+            return ENOMEM;
+        }
+        __sync_bool_compare_and_swap(&group->alives, NULL, sptr_array);
+    }
+
+    return 0;
+}
+
+int sf_connection_manager_start(SFConnectionManager *cm)
+{
+    pthread_t tid;
+
+    if (cm->alive_detect.bg_thread_enabled) {
+        return fc_create_thread(&tid, connection_manager_thread_func,
+                cm, SF_G_THREAD_STACK_SIZE);
+    } else {
+        return 0;
+    }
+}

src/sf_connection_manager.h

@@ -0,0 +1,173 @@
+/*
+ * Copyright (c) 2020 YuQing <384681@qq.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <https://www.gnu.org/licenses/>.
+ */
+
+//sf_connection_manager.h
+
+#ifndef _SF_CONNECTION_MANAGER_H
+#define _SF_CONNECTION_MANAGER_H
+
+#include "fastcommon/common_blocked_queue.h"
+#include "fastcommon/server_id_func.h"
+#include "fastcommon/connection_pool.h"
+#include "sf_types.h"
+
+struct sf_connection_manager;
+
+typedef ConnectionInfo *(*sf_get_connection_func)(
+        struct sf_connection_manager *cm, const int group_index,
+        const bool shared, int *err_no);
+
+typedef ConnectionInfo *(*sf_get_server_connection_func)(
+        struct sf_connection_manager *cm,
+        FCServerInfo *server, const bool shared, int *err_no);
+
+typedef ConnectionInfo *(*sf_get_spec_connection_func)(
+        struct sf_connection_manager *cm,
+        const ConnectionInfo *target,
+        const bool shared, int *err_no);
+
+typedef void (*sf_release_connection_func)(
+        struct sf_connection_manager *cm, ConnectionInfo *conn);
+typedef void (*sf_close_connection_func)(
+        struct sf_connection_manager *cm, ConnectionInfo *conn);
+
+typedef const struct sf_connection_parameters * (*sf_get_connection_parameters)(
+        struct sf_connection_manager *cm, ConnectionInfo *conn);
+
+typedef struct sf_cm_server_entry {
+    int id;
+    int group_index;
+    FCAddressPtrArray *addr_array;
+} SFCMServerEntry;
+
+typedef struct sf_cm_server_array {
+    SFCMServerEntry *servers;
+    int count;
+} SFCMServerArray;
+
+typedef struct sf_cm_server_ptr_array {
+    SFCMServerEntry **servers;
+    int count;
+} SFCMServerPtrArray;
+
+typedef struct sf_cm_conn_group_entry {
+    int id;
+    volatile char in_queue;  //if in active detect queue
+    SFCMServerArray all;
+    volatile SFCMServerEntry *master;
+    volatile SFCMServerPtrArray *alives;
+} SFCMConnGroupEntry;
+
+typedef struct sf_cm_conn_group_array {
+    SFCMConnGroupEntry *entries;
+    int count;
+} SFCMConnGroupArray;
+
+typedef struct sf_cm_operations {
+    /* get the specify connection by ip and port */
+    sf_get_spec_connection_func get_spec_connection;
+
+    /* get one connection of the configured servers by data group */
+    sf_get_connection_func get_connection;
+
+    /* get one connection of the server */
+    sf_get_server_connection_func get_server_connection;
+
+    /* get the master connection from the server */
+    sf_get_connection_func get_master_connection;
+
+    /* get one readable connection from the server */
+    sf_get_connection_func get_readable_connection;
+
+    /* get the leader connection from the server */
+    sf_get_server_connection_func get_leader_connection;
+
+    /* push back to connection pool when use connection pool */
+    sf_release_connection_func release_connection;
+
+     /* disconnect the connecton on network error */
+    sf_close_connection_func close_connection;
+
+    sf_get_connection_parameters get_connection_params;
+} SFCMOperations;
+
+typedef struct sf_connection_manager {
+    int exclude_server_id;  //for server side
+    uint16_t max_servers_per_group;
+    uint8_t server_group_index;
+    struct {
+        bool bg_thread_enabled;
+        struct common_blocked_queue queue;
+    } alive_detect;
+    const char *module_name;
+    const SFClientCommonConfig *common_cfg;
+    SFCMConnGroupArray groups;
+    ConnectionPool cpool;
+    struct fast_mblock_man sptr_array_allocator; //element: SFCMServerPtrArray
+    SFCMOperations ops;
+    FCServerConfig *server_cfg;
+    void *extra;   //for simple connection manager
+} SFConnectionManager;
+
+int sf_connection_manager_init_ex(SFConnectionManager *cm,
+        const char *module_name, const SFClientCommonConfig *common_cfg,
+        const int group_count, const int server_group_index,
+        const int server_count, const int max_count_per_entry,
+        const int max_idle_time, fc_connection_callback_func
+        connect_done_callback, void *args, FCServerConfig *server_cfg,
+        const bool bg_thread_enabled);
+
+static inline int sf_connection_manager_init(SFConnectionManager *cm,
+        const char *module_name, const SFClientCommonConfig *common_cfg,
+        const int group_count, const int server_group_index,
+        const int server_count, const int max_count_per_entry,
+        const int max_idle_time, fc_connection_callback_func
+        connect_done_callback, void *args)
+{
+    const bool bg_thread_enabled = true;
+    return sf_connection_manager_init_ex(cm, module_name,
+            common_cfg, group_count, server_group_index,
+            server_count, max_count_per_entry, max_idle_time,
+            connect_done_callback, args, NULL, bg_thread_enabled);
+}
+
+static inline void sf_connection_manager_set_exclude_server_id(
+        SFConnectionManager *cm, const int exclude_server_id)
+{
+    cm->exclude_server_id = exclude_server_id;
+}
+
+int sf_connection_manager_add(SFConnectionManager *cm, const int group_id,
+        FCServerInfo **servers, const int count);
+
+int sf_connection_manager_prepare(SFConnectionManager *cm);
+
+//start thread
+int sf_connection_manager_start(SFConnectionManager *cm);
+
+int sf_cm_validate_connection_callback(ConnectionInfo *conn, void *args);
+
+const struct sf_connection_parameters *sf_cm_get_connection_params(
+        SFConnectionManager *cm, ConnectionInfo *conn);
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif

src/sf_define.h

@@ -20,13 +20,17 @@
 
 #include "fastcommon/sockopt.h"
 
-#define SF_DEF_THREAD_STACK_SIZE  (64 * 1024)
+#define SF_DEFAULT_CONNECT_TIMEOUT   10
+#define SF_DEFAULT_NETWORK_TIMEOUT   60
+
+#define SF_DEF_THREAD_STACK_SIZE  (256 * 1024)
 #define SF_MIN_THREAD_STACK_SIZE  (64 * 1024)
 #define SF_MAX_THREAD_STACK_SIZE  (2 * 1024 * 1024 * 1024LL)
-#define SF_DEF_MAX_PACKAGE_SIZE   (64 * 1024)
+#define SF_DEF_MAX_PACKAGE_SIZE   (256 * 1024)
 #define SF_DEF_MIN_BUFF_SIZE      (64 * 1024)
-#define SF_DEF_MAX_BUFF_SIZE      (64 * 1024)
+#define SF_DEF_MAX_BUFF_SIZE      (256 * 1024)
 #define SF_MAX_NETWORK_BUFF_SIZE  (2 * 1024 * 1024 * 1024LL)
+#define SF_DEF_SYNC_LOG_BUFF_INTERVAL  1
 
 #define SF_NIO_STAGE_NONE        0
 #define SF_NIO_STAGE_INIT        1  //set ioevent
@@ -40,6 +44,10 @@
 
 #define SF_NIO_TASK_STAGE_FETCH(task)  task->nio_stages.current
 
+#define SF_SESSION_ERROR_NOT_EXIST           9992
+#define SF_CLUSTER_ERROR_NOT_LEADER          9995
+#define SF_CLUSTER_ERROR_LEADER_VERSION_INCONSISTENT 9996
+#define SF_CLUSTER_ERROR_BINLOG_MISSED       9997
 #define SF_CLUSTER_ERROR_BINLOG_INCONSISTENT 9998
 #define SF_CLUSTER_ERROR_LEADER_INCONSISTENT 9999
 #define SF_CLUSTER_ERROR_MASTER_INCONSISTENT SF_CLUSTER_ERROR_LEADER_INCONSISTENT
@@ -53,9 +61,25 @@
 #define SF_RETRIABLE_ERROR_NO_CHANNEL      9914
 #define SF_RETRIABLE_ERROR_CHANNEL_INVALID 9915  //client should re-setup channel
 
+//std errno wrapper for crossing platform
+#define SF_ERROR_EBUSY        8816
+#define SF_ERROR_EINVAL       8822
+#define SF_ERROR_EAGAIN       8835
+#define SF_ERROR_EINPROGRESS  8836
+#define SF_ERROR_EOVERFLOW    8884
+#define SF_ERROR_EOPNOTSUPP   8895
+#define SF_ERROR_ENOLINK      8867
+#define SF_ERROR_ENODATA      8861
+#define SF_ERROR_ENOTEMPTY    8839
+#define SF_ERROR_ELOOP        8840
+
 #define SF_FORCE_CLOSE_CONNECTION_ERROR_MIN  SF_RETRIABLE_ERROR_NOT_MASTER
 #define SF_FORCE_CLOSE_CONNECTION_ERROR_MAX  SF_RETRIABLE_ERROR_MAX
 
+#define SF_IS_SERVER_RETRIABLE_ERROR(code) \
+    ((code >= SF_RETRIABLE_ERROR_MIN && code <= SF_RETRIABLE_ERROR_MAX) || \
+     (code == EAGAIN))
+
 #define SF_IS_RETRIABLE_ERROR(code) \
     ((code >= SF_RETRIABLE_ERROR_MIN && code <= SF_RETRIABLE_ERROR_MAX) || \
      (code == EAGAIN) || is_network_error(code))
@@ -66,13 +90,14 @@
       (result == EINVAL) || (result == EOVERFLOW)  || \
       (result != 0 && is_network_error(code)))
 
-
 #define SF_UNIX_ERRNO(code, errno_for_overflow) \
     (code < 256 ? code : errno_for_overflow)
 
 #define SF_BINLOG_SOURCE_USER        'U'  //by user call
 #define SF_BINLOG_SOURCE_REPLAY      'R'  //by binlog replay
 
+#define SF_LOG_SCHEDULE_ENTRIES_COUNT 3
+
 #ifdef __cplusplus
 extern "C" {
 #endif

src/sf_file_writer.c

@@ -0,0 +1,707 @@
+/*
+ * Copyright (c) 2020 YuQing <384681@qq.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <https://www.gnu.org/licenses/>.
+ */
+
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <sys/socket.h>
+#include <netinet/in.h>
+#include <arpa/inet.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <errno.h>
+#include <limits.h>
+#include <fcntl.h>
+#include <pthread.h>
+#include "fastcommon/logger.h"
+#include "fastcommon/sockopt.h"
+#include "fastcommon/shared_func.h"
+#include "fastcommon/pthread_func.h"
+#include "fastcommon/sched_thread.h"
+#include "sf_global.h"
+#include "sf_func.h"
+#include "sf_file_writer.h"
+
+#define BINLOG_INDEX_ITEM_START_INDEX_STR       "start_index"
+#define BINLOG_INDEX_ITEM_START_INDEX_LEN      \
+    (sizeof(BINLOG_INDEX_ITEM_START_INDEX_STR) - 1)
+
+#define BINLOG_INDEX_ITEM_CURRENT_WRITE_STR     "current_write"
+#define BINLOG_INDEX_ITEM_CURRENT_WRITE_LEN      \
+    (sizeof(BINLOG_INDEX_ITEM_CURRENT_WRITE_STR) - 1)
+
+#define BINLOG_INDEX_ITEM_CURRENT_COMPRESS_STR  "current_compress"
+#define BINLOG_INDEX_ITEM_CURRENT_COMPRESS_LEN      \
+    (sizeof(BINLOG_INDEX_ITEM_CURRENT_COMPRESS_STR) - 1)
+
+static inline void sf_file_writer_get_binlog_filename(SFFileWriterInfo *writer)
+{
+    sf_file_writer_get_filename_ex(
+            writer->cfg.data_path, writer->cfg.subdir_name,
+            writer->cfg.file_prefix, writer->binlog.last_index,
+            writer->file.name.str, writer->file.name.size);
+}
+
+static inline void sf_file_writer_get_index_filename_ex(
+        const char *data_path, const char *subdir_name,
+        const char *file_prefix, const int file_prefix_len,
+        char *filename, const int size)
+{
+#define INDEX_FILENAME_AFFIX_STR "_index.dat"
+#define INDEX_FILENAME_AFFIX_LEN (sizeof(INDEX_FILENAME_AFFIX_STR) - 1)
+
+    char *p;
+    int data_path_len;
+    int subdir_name_len;
+
+    data_path_len = strlen(data_path);
+    subdir_name_len = strlen(subdir_name);
+    if (data_path_len + 1 + subdir_name_len + 1 + file_prefix_len +
+            INDEX_FILENAME_AFFIX_LEN >= size)
+    {
+        *filename = '\0';
+        return;
+    }
+
+    memcpy(filename, data_path, data_path_len);
+    p = filename + data_path_len;
+    *p++ = '/';
+
+    memcpy(p, subdir_name, subdir_name_len);
+    p += subdir_name_len;
+    *p++ = '/';
+
+    memcpy(p, file_prefix, file_prefix_len);
+    p += file_prefix_len;
+
+    memcpy(p, INDEX_FILENAME_AFFIX_STR, INDEX_FILENAME_AFFIX_LEN);
+    p += INDEX_FILENAME_AFFIX_LEN;
+    *p = '\0';
+}
+
+const char *sf_file_writer_get_index_filename(const char *data_path,
+        const char *subdir_name, char *filename, const int size)
+{
+    sf_file_writer_get_index_filename_ex(data_path, subdir_name,
+            SF_BINLOG_FILE_PREFIX_STR, SF_BINLOG_FILE_PREFIX_LEN,
+            filename, size);
+    return filename;
+}
+
+int sf_file_writer_write_to_binlog_index_file_ex(const char *data_path,
+        const char *subdir_name, const char *file_prefix,
+        const int start_index, const int last_index,
+        const int compress_index)
+{
+    char filename[PATH_MAX];
+    char buff[256];
+    char *p;
+    int result;
+    int len;
+
+    sf_file_writer_get_index_filename_ex(data_path, subdir_name, file_prefix,
+            strlen(file_prefix), filename, sizeof(filename));
+    p = buff;
+    memcpy(p, BINLOG_INDEX_ITEM_START_INDEX_STR,
+            BINLOG_INDEX_ITEM_START_INDEX_LEN);
+    p += BINLOG_INDEX_ITEM_START_INDEX_LEN;
+    *p++ = '=';
+    p += fc_itoa(start_index, p);
+    *p++ = '\n';
+
+    memcpy(p, BINLOG_INDEX_ITEM_CURRENT_WRITE_STR,
+            BINLOG_INDEX_ITEM_CURRENT_WRITE_LEN);
+    p += BINLOG_INDEX_ITEM_CURRENT_WRITE_LEN;
+    *p++ = '=';
+    p += fc_itoa(last_index, p);
+    *p++ = '\n';
+
+    memcpy(p, BINLOG_INDEX_ITEM_CURRENT_COMPRESS_STR,
+            BINLOG_INDEX_ITEM_CURRENT_COMPRESS_LEN);
+    p += BINLOG_INDEX_ITEM_CURRENT_COMPRESS_LEN;
+    *p++ = '=';
+    p += fc_itoa(compress_index, p);
+    *p++ = '\n';
+
+    len = p - buff;
+    if ((result=safeWriteToFile(filename, buff, len)) != 0) {
+        logError("file: "__FILE__", line: %d, "
+                "write to file \"%s\" fail, errno: %d, error info: %s",
+                __LINE__, filename, result, STRERROR(result));
+    }
+
+    return result;
+}
+
+static inline int write_to_binlog_index_file(SFFileWriterInfo *writer)
+{
+    return sf_file_writer_write_to_binlog_index_file_ex(
+            writer->cfg.data_path, writer->cfg.subdir_name,
+            writer->cfg.file_prefix, writer->binlog.start_index,
+            writer->binlog.last_index, writer->binlog.compress_index);
+}
+
+static int get_binlog_info_from_file(const char *data_path,
+        const char *subdir_name, int *start_index,
+        int *last_index, int *compress_index)
+{
+    char full_filename[PATH_MAX];
+    IniContext ini_context;
+    int result;
+
+    sf_file_writer_get_index_filename_ex(data_path, subdir_name,
+            SF_BINLOG_FILE_PREFIX_STR, SF_BINLOG_FILE_PREFIX_LEN,
+            full_filename, sizeof(full_filename));
+    if (access(full_filename, F_OK) != 0) {
+        return errno != 0 ? errno : EPERM;
+    }
+
+    if ((result=iniLoadFromFile(full_filename, &ini_context)) != 0) {
+        logError("file: "__FILE__", line: %d, "
+                "load from file \"%s\" fail, error code: %d",
+                __LINE__, full_filename, result);
+        return result;
+    }
+
+    *start_index = iniGetIntValue(NULL,
+            BINLOG_INDEX_ITEM_START_INDEX_STR,
+            &ini_context, 0);
+    *last_index = iniGetIntValue(NULL,
+            BINLOG_INDEX_ITEM_CURRENT_WRITE_STR,
+            &ini_context, 0);
+    *compress_index = iniGetIntValue(NULL,
+            BINLOG_INDEX_ITEM_CURRENT_COMPRESS_STR,
+            &ini_context, 0);
+
+    iniFreeContext(&ini_context);
+    return 0;
+}
+
+int sf_file_writer_get_binlog_indexes(const char *data_path,
+        const char *subdir_name, int *start_index, int *last_index)
+{
+    int result;
+    int compress_index;
+
+    result = get_binlog_info_from_file(data_path, subdir_name,
+            start_index, last_index, &compress_index);
+    if (result == ENOENT) {
+        *start_index = *last_index = 0;
+        return 0;
+    } else {
+        return result;
+    }
+}
+
+static inline int get_binlog_index_from_file(SFFileWriterInfo *writer)
+{
+    int result;
+
+    result = get_binlog_info_from_file(writer->cfg.data_path,
+            writer->cfg.subdir_name, &writer->binlog.start_index,
+            &writer->binlog.last_index, &writer->binlog.compress_index);
+    if (result == ENOENT) {
+        writer->binlog.start_index = 0;
+        writer->binlog.last_index = 0;
+        writer->binlog.compress_index = 0;
+        if (writer->cfg.file_rotate_size > 0) {
+            return write_to_binlog_index_file(writer);
+        } else {
+            return 0;
+        }
+    } else {
+        return result;
+    }
+}
+
+static int open_writable_binlog(SFFileWriterInfo *writer)
+{
+    if (writer->file.fd >= 0) {
+        close(writer->file.fd);
+    }
+
+    sf_file_writer_get_binlog_filename(writer);
+    writer->file.fd = open(writer->file.name.str, O_WRONLY |
+            O_CREAT | O_APPEND | O_CLOEXEC, 0644);
+    if (writer->file.fd < 0) {
+        logError("file: "__FILE__", line: %d, "
+                "open file \"%s\" fail, "
+                "errno: %d, error info: %s",
+                __LINE__, writer->file.name.str,
+                errno, STRERROR(errno));
+        return errno != 0 ? errno : EACCES;
+    }
+
+    writer->file.size = lseek(writer->file.fd, 0, SEEK_END);
+    if (writer->file.size < 0) {
+        logError("file: "__FILE__", line: %d, "
+                "lseek file \"%s\" fail, "
+                "errno: %d, error info: %s",
+                __LINE__, writer->file.name.str,
+                errno, STRERROR(errno));
+        return errno != 0 ? errno : EIO;
+    }
+
+    return 0;
+}
+
+static int open_next_binlog(SFFileWriterInfo *writer)
+{
+    sf_file_writer_get_binlog_filename(writer);
+    if (access(writer->file.name.str, F_OK) == 0) {
+        char bak_filename[PATH_MAX];
+        char date_str[32];
+
+        formatDatetime(g_current_time, "%Y%m%d%H%M%S",
+                date_str, sizeof(date_str));
+        fc_combine_two_strings(writer->file.name.str,
+                date_str, '.', bak_filename);
+        if (rename(writer->file.name.str, bak_filename) == 0) {
+            logWarning("file: "__FILE__", line: %d, "
+                    "binlog file %s exist, rename to %s",
+                    __LINE__, writer->file.name.str, bak_filename);
+        } else {
+            logError("file: "__FILE__", line: %d, "
+                    "rename binlog %s to backup %s fail, "
+                    "errno: %d, error info: %s",
+                    __LINE__, writer->file.name.str, bak_filename,
+                    errno, STRERROR(errno));
+            return errno != 0 ? errno : EPERM;
+        }
+    }
+
+    return open_writable_binlog(writer);
+}
+
+static int do_write_to_file(SFFileWriterInfo *writer,
+        char *buff, const int len)
+{
+    int result;
+
+    if (fc_safe_write(writer->file.fd, buff, len) != len) {
+        result = errno != 0 ? errno : EIO;
+        logError("file: "__FILE__", line: %d, "
+                "write to binlog file \"%s\" fail, "
+                "errno: %d, error info: %s",
+                __LINE__, writer->file.name.str,
+                result, STRERROR(result));
+        return result;
+    }
+
+    if (writer->cfg.call_fsync) {
+        if (fsync(writer->file.fd) != 0) {
+            result = errno != 0 ? errno : EIO;
+            logError("file: "__FILE__", line: %d, "
+                    "fsync to binlog file \"%s\" fail, errno: %d, "
+                    "error info: %s", __LINE__, writer->file.name.str,
+                    result, STRERROR(result));
+            return result;
+        }
+    }
+
+    writer->file.size += len;
+    if (writer->write_done_callback.func != NULL) {
+        writer->write_done_callback.func(writer,
+                writer->write_done_callback.args);
+    }
+
+    return 0;
+}
+
+int sf_file_writer_direct_write(SFFileWriterInfo *writer,
+        char *buff, const int len)
+{
+    int result;
+
+    if ((writer->cfg.file_rotate_size <= 0) || (writer->file.size
+                + len <= writer->cfg.file_rotate_size))
+    {
+        return do_write_to_file(writer, buff, len);
+    }
+
+    writer->binlog.last_index++;  //binlog rotate
+    if ((result=write_to_binlog_index_file(writer)) == 0) {
+        result = open_next_binlog(writer);
+    }
+
+    if (result != 0) {
+        logError("file: "__FILE__", line: %d, "
+                "open binlog file \"%s\" fail",
+                __LINE__, writer->file.name.str);
+        return result;
+    }
+
+    return do_write_to_file(writer, buff, len);
+}
+
+int sf_file_writer_flush(SFFileWriterInfo *writer)
+{
+    int result;
+    int len;
+
+    len = SF_BINLOG_BUFFER_PRODUCER_DATA_LENGTH(writer->binlog_buffer);
+    if (len == 0) {
+        return 0;
+    }
+
+    if ((result=sf_file_writer_direct_write(writer, writer->
+                    binlog_buffer.buff, len)) == 0)
+    {
+        if (writer->flags & SF_FILE_WRITER_FLAGS_WANT_DONE_VERSION) {
+            writer->last_versions.done = writer->last_versions.pending;
+        }
+    }
+
+    writer->binlog_buffer.data_end = writer->binlog_buffer.buff;
+    return result;
+}
+
+int sf_file_writer_fsync(SFFileWriterInfo *writer)
+{
+    int result;
+
+    if ((result=sf_file_writer_flush(writer)) != 0) {
+        return result;
+    }
+
+    if (fsync(writer->file.fd) == 0) {
+        return 0;
+    } else {
+        result = errno != 0 ? errno : EIO;
+        logError("file: "__FILE__", line: %d, "
+                "fsync to binlog file \"%s\" fail, errno: %d, "
+                "error info: %s", __LINE__, writer->file.name.str,
+                result, STRERROR(result));
+        return result;
+    }
+}
+
+int sf_file_writer_get_indexes(SFFileWriterInfo *writer,
+        int *start_index, int *last_index)
+{
+    int result;
+
+    if (writer == NULL) {   //for data recovery
+        *start_index = *last_index = 0;
+        return 0;
+    }
+
+    if (writer->binlog.last_index < 0) {
+        if ((result=get_binlog_index_from_file(writer)) != 0) {
+            *start_index = *last_index = -1;
+            return result;
+        }
+    }
+
+    *start_index = writer->binlog.start_index;
+    *last_index = writer->binlog.last_index;
+    return 0;
+}
+
+int sf_file_writer_deal_versioned_buffer(SFFileWriterInfo *writer,
+        BufferInfo *buffer, const int64_t version)
+{
+    int result;
+
+    if (buffer->length >= writer->binlog_buffer.size / 4) {
+        if (SF_BINLOG_BUFFER_PRODUCER_DATA_LENGTH(writer->binlog_buffer) > 0) {
+            if ((result=sf_file_writer_flush(writer)) != 0) {
+                return result;
+            }
+        }
+
+        if ((result=sf_file_writer_direct_write(writer, buffer->
+                        buff, buffer->length)) == 0)
+        {
+            if (writer->flags & SF_FILE_WRITER_FLAGS_WANT_DONE_VERSION) {
+                writer->last_versions.pending = version;
+                writer->last_versions.done = version;
+            }
+        }
+        return result;
+    }
+
+    if (writer->cfg.file_rotate_size > 0 && writer->file.size +
+            SF_BINLOG_BUFFER_PRODUCER_DATA_LENGTH(writer->binlog_buffer) +
+            buffer->length > writer->cfg.file_rotate_size)
+    {
+        if ((result=sf_file_writer_flush(writer)) != 0) {
+            return result;
+        }
+    } else if (SF_BINLOG_BUFFER_PRODUCER_BUFF_REMAIN(
+                writer->binlog_buffer) < buffer->length)
+    {
+        if ((result=sf_file_writer_flush(writer)) != 0) {
+            return result;
+        }
+    }
+
+    if (writer->flags & SF_FILE_WRITER_FLAGS_WANT_DONE_VERSION) {
+        writer->last_versions.pending = version;
+    }
+    memcpy(writer->binlog_buffer.data_end, buffer->buff, buffer->length);
+    writer->binlog_buffer.data_end += buffer->length;
+
+    return 0;
+}
+
+int sf_file_writer_save_buffer_ex(SFFileWriterInfo *writer,
+        const int length, const bool flush)
+{
+    int result;
+
+    if (writer->cfg.file_rotate_size > 0 && writer->file.size +
+            SF_BINLOG_BUFFER_PRODUCER_DATA_LENGTH(writer->binlog_buffer) +
+            length > writer->cfg.file_rotate_size)
+    {
+        if ((result=sf_file_writer_flush(writer)) != 0) {
+            return result;
+        }
+    }
+
+    writer->binlog_buffer.data_end += length;
+    if (flush || SF_BINLOG_BUFFER_PRODUCER_BUFF_REMAIN(writer->
+                binlog_buffer) < writer->cfg.max_record_size)
+    {
+        return sf_file_writer_flush(writer);
+    } else {
+        return 0;
+    }
+}
+
+int sf_file_writer_init(SFFileWriterInfo *writer, const char *data_path,
+        const char *subdir_name, const char *file_prefix,
+        const int max_record_size, const int buffer_size,
+        const int64_t file_rotate_size, const bool call_fsync)
+{
+    int result;
+    int path_len;
+    bool create;
+    char filepath[PATH_MAX];
+
+    writer->total_count = 0;
+    writer->last_versions.pending = 0;
+    writer->last_versions.done = 0;
+    writer->flags = 0;
+    sf_file_writer_set_write_done_callback(writer, NULL, NULL);
+    if ((result=sf_binlog_buffer_init(&writer->
+                    binlog_buffer, buffer_size)) != 0)
+    {
+        return result;
+    }
+
+    writer->cfg.max_record_size = max_record_size;
+    writer->cfg.call_fsync = call_fsync;
+    writer->cfg.file_rotate_size = file_rotate_size;
+    writer->cfg.data_path = data_path;
+    path_len = fc_combine_full_filepath(data_path, subdir_name, filepath);
+    if ((result=fc_check_mkdir_ex(filepath, 0775, &create)) != 0) {
+        return result;
+    }
+    if (create) {
+        SF_CHOWN_TO_RUNBY_RETURN_ON_ERROR(filepath);
+    }
+
+    writer->file.fd = -1;
+    fc_safe_strcpy(writer->cfg.subdir_name, subdir_name);
+    fc_safe_strcpy(writer->cfg.file_prefix, file_prefix);
+    writer->file.name.size = path_len + 32;
+    writer->file.name.str = (char *)fc_malloc(writer->file.name.size);
+    if (writer->file.name.str == NULL) {
+        return ENOMEM;
+    }
+
+    if ((result=get_binlog_index_from_file(writer)) != 0) {
+        return result;
+    }
+
+    if ((result=open_writable_binlog(writer)) != 0) {
+        return result;
+    }
+
+    return 0;
+}
+
+void sf_file_writer_destroy(SFFileWriterInfo *writer)
+{
+    if (writer->file.fd >= 0) {
+        close(writer->file.fd);
+        writer->file.fd = -1;
+    }
+    if (writer->file.name.str != NULL) {
+        free(writer->file.name.str);
+        writer->file.name.str = NULL;
+    }
+    sf_binlog_buffer_destroy(&writer->binlog_buffer);
+}
+
+int sf_file_writer_set_indexes(SFFileWriterInfo *writer,
+        const int start_index, const int last_index)
+{
+    int result;
+
+    if (writer->binlog.start_index != start_index ||
+            writer->binlog.last_index != last_index)
+    {
+        writer->binlog.start_index = start_index;
+        writer->binlog.last_index = last_index;
+        if ((result=write_to_binlog_index_file(writer)) != 0) {
+            return result;
+        }
+    }
+
+    return 0;
+}
+
+int sf_file_writer_set_binlog_start_index(SFFileWriterInfo *writer,
+        const int start_index)
+{
+    int result;
+
+    if (writer->binlog.start_index != start_index) {
+        writer->binlog.start_index = start_index;
+        if ((result=write_to_binlog_index_file(writer)) != 0) {
+            return result;
+        }
+    }
+
+    return 0;
+}
+
+int sf_file_writer_set_binlog_write_index(SFFileWriterInfo *writer,
+        const int last_index)
+{
+    int result;
+
+    if (writer->binlog.last_index != last_index) {
+        writer->binlog.last_index = last_index;
+        if ((result=write_to_binlog_index_file(writer)) != 0) {
+            return result;
+        }
+    }
+
+    return open_writable_binlog(writer);
+}
+
+int sf_file_writer_get_last_lines(const char *data_path,
+        const char *subdir_name, const int current_write_index,
+        char *buff, const int buff_size, int *count, int *length)
+{
+    int result;
+    int target_count;
+    int count1;
+    char filename[PATH_MAX];
+    string_t lines;
+
+    target_count = *count;
+    sf_file_writer_get_filename(data_path, subdir_name,
+            current_write_index, filename, sizeof(filename));
+    if (access(filename, F_OK) == 0) {
+        if ((result=fc_get_last_lines(filename, buff, buff_size,
+                        &lines, count)) != 0)
+        {
+            if (result != ENOENT) {
+                return result;
+            }
+        }
+
+        if (*count >= target_count || current_write_index == 0) {
+            memmove(buff, lines.str, lines.len);
+            *length = lines.len;
+            return 0;
+        }
+    } else {
+        result = errno != 0 ? errno : EPERM;
+        if (result == ENOENT) {
+            *count = 0;
+            *length = 0;
+            return 0;
+        } else {
+            logError("file: "__FILE__", line: %d, "
+                    "stat file %s fail, errno: %d, error info: %s",
+                    __LINE__, filename, result, STRERROR(result));
+            *count = 0;
+            *length = 0;
+            return result;
+        }
+    }
+
+    sf_file_writer_get_filename(data_path, subdir_name,
+            current_write_index - 1, filename, sizeof(filename));
+    if (access(filename, F_OK) != 0) {
+        result = errno != 0 ? errno : EPERM;
+        if (result == ENOENT) {
+            memmove(buff, lines.str, lines.len);
+            *length = lines.len;
+            return 0;
+        } else {
+            logError("file: "__FILE__", line: %d, "
+                    "stat file %s fail, errno: %d, error info: %s",
+                    __LINE__, filename, result, STRERROR(result));
+            *count = 0;
+            *length = 0;
+            return result;
+        }
+    }
+
+    count1 = target_count - *count;
+    if ((result=fc_get_last_lines(filename, buff,
+                    buff_size, &lines, &count1)) != 0)
+    {
+        *count = 0;
+        *length = 0;
+        return result;
+    }
+
+    memmove(buff, lines.str, lines.len);
+    *length = lines.len;
+    if (*count == 0) {
+        *count = count1;
+    } else {
+        sf_file_writer_get_filename(data_path, subdir_name,
+                current_write_index, filename, sizeof(filename));
+        if ((result=fc_get_first_lines(filename, buff + (*length),
+                        buff_size - (*length), &lines, count)) != 0)
+        {
+            *count = 0;
+            *length = 0;
+            return result;
+        }
+
+        *count += count1;
+        *length += lines.len;
+    }
+
+    return 0;
+}
+
+int sf_file_writer_get_last_line(const char *data_path,
+        const char *subdir_name, char *buff,
+        const int buff_size, int *length)
+{
+    int result;
+    int last_index;
+    int count = 1;
+
+    if ((result=sf_file_writer_get_binlog_last_index(data_path,
+                    subdir_name, &last_index)) != 0)
+    {
+        *length = 0;
+        return result;
+    }
+
+    return sf_file_writer_get_last_lines(data_path, subdir_name,
+            last_index, buff, buff_size, &count, length);
+}

src/sf_file_writer.h

@@ -0,0 +1,323 @@
+/*
+ * Copyright (c) 2020 YuQing <384681@qq.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <https://www.gnu.org/licenses/>.
+ */
+
+//sf_file_writer.h
+
+#ifndef _SF_FILE_WRITER_H_
+#define _SF_FILE_WRITER_H_
+
+#include "fastcommon/fc_queue.h"
+#include "sf_types.h"
+
+#define SF_FILE_WRITER_FLAGS_WANT_DONE_VERSION  1
+
+#define SF_BINLOG_SUBDIR_NAME_SIZE     128
+#define SF_BINLOG_FILE_PREFIX_STR_SIZE 64
+#define SF_BINLOG_DEFAULT_ROTATE_SIZE  (1024 * 1024 * 1024)
+#define SF_BINLOG_NEVER_ROTATE_FILE    0
+#define SF_BINLOG_FILE_PREFIX_STR      "binlog"
+#define SF_BINLOG_FILE_PREFIX_LEN      (sizeof(SF_BINLOG_FILE_PREFIX_STR) - 1)
+#define SF_BINLOG_FILE_EXT_LEN 6
+#define SF_BINLOG_FILE_EXT_FMT ".%0"FC_MACRO_TOSTRING(SF_BINLOG_FILE_EXT_LEN)"d"
+
+struct sf_file_writer_info;
+
+typedef void (*sf_file_write_done_callback)(
+        struct sf_file_writer_info *writer, void *args);
+
+typedef struct sf_file_writer_info {
+    struct {
+        const char *data_path;
+        char subdir_name[SF_BINLOG_SUBDIR_NAME_SIZE];
+        char file_prefix[SF_BINLOG_FILE_PREFIX_STR_SIZE];
+        int64_t file_rotate_size;
+        int max_record_size;
+        bool call_fsync;
+    } cfg;
+
+    struct {
+        int start_index;  //for read only
+        int last_index;   //for write
+        int compress_index;
+    } binlog;
+
+    struct {
+        int fd;
+        int64_t size;  //file size
+        struct {
+            char *str;
+            int size;
+        } name;
+    } file;
+
+    int64_t total_count;
+    SFBinlogBuffer binlog_buffer;
+
+    short flags;
+    struct {
+        int64_t pending;
+        volatile int64_t done;
+    } last_versions;
+
+    struct {
+        sf_file_write_done_callback func;
+        void *args;
+    } write_done_callback;
+
+} SFFileWriterInfo;
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+int sf_file_writer_init(SFFileWriterInfo *writer, const char *data_path,
+        const char *subdir_name, const char *file_prefix,
+        const int max_record_size, const int buffer_size,
+        const int64_t file_rotate_size, const bool call_fsync);
+
+void sf_file_writer_destroy(SFFileWriterInfo *writer);
+
+int sf_file_writer_direct_write(SFFileWriterInfo *writer,
+        char *buff, const int len);
+
+int sf_file_writer_deal_versioned_buffer(SFFileWriterInfo *writer,
+        BufferInfo *buffer, const int64_t version);
+
+#define sf_file_writer_deal_buffer(writer, buffer) \
+    sf_file_writer_deal_versioned_buffer(writer, buffer, 0)
+
+int sf_file_writer_flush(SFFileWriterInfo *writer);
+
+int sf_file_writer_fsync(SFFileWriterInfo *writer);
+
+#define SF_FILE_WRITER_DATA_END_BUFF(writer) (writer)->binlog_buffer.data_end
+#define SF_FILE_WRITER_CURRENT_DATA_VERSION(writer) \
+    (writer)->last_versions.pending
+#define SF_FILE_WRITER_NEXT_DATA_VERSION(writer) \
+    ++((writer)->last_versions.pending)
+
+int sf_file_writer_save_buffer_ex(SFFileWriterInfo *writer,
+        const int length, const bool flush);
+
+static inline int sf_file_writer_save_buffer(
+        SFFileWriterInfo *writer, const int length)
+{
+    const bool flush = false;
+    return sf_file_writer_save_buffer_ex(writer, length, flush);
+}
+
+static inline int sf_file_writer_flush_buffer(
+        SFFileWriterInfo *writer, const int length)
+{
+    const bool flush = true;
+    return sf_file_writer_save_buffer_ex(writer, length, flush);
+}
+
+static inline void sf_file_writer_set_flags(
+        SFFileWriterInfo *writer, const short flags)
+{
+    writer->flags = flags;
+}
+
+static inline void sf_file_writer_set_call_fsync(
+        SFFileWriterInfo *writer, const bool call_fsync)
+{
+    writer->cfg.call_fsync = call_fsync;
+}
+
+static inline void sf_file_writer_set_write_done_callback (
+        SFFileWriterInfo *writer, sf_file_write_done_callback callback,
+        void *args)
+{
+    writer->write_done_callback.func = callback;
+    writer->write_done_callback.args = args;
+}
+
+static inline int64_t sf_file_writer_get_last_version_ex(
+        SFFileWriterInfo *writer, const int log_level)
+{
+    if (writer->flags & SF_FILE_WRITER_FLAGS_WANT_DONE_VERSION) {
+        return writer->last_versions.done;
+    } else {
+        if (FC_LOG_BY_LEVEL(log_level)) {
+            log_it_ex(&g_log_context, log_level, "file: %s, line: %d, "
+                    "writer: %s, should set writer flags to %d!",
+                    __FILE__, __LINE__, writer->cfg.subdir_name,
+                    SF_FILE_WRITER_FLAGS_WANT_DONE_VERSION);
+        }
+
+        return -1;
+    }
+}
+
+#define sf_file_writer_get_last_version(writer) \
+    sf_file_writer_get_last_version_ex(writer, LOG_ERR)
+
+#define sf_file_writer_get_last_version_silence(writer) \
+    sf_file_writer_get_last_version_ex(writer, LOG_NOTHING)
+
+int sf_file_writer_get_binlog_indexes(const char *data_path,
+        const char *subdir_name, int *start_index, int *last_index);
+
+static inline int sf_file_writer_get_binlog_start_index(
+        const char *data_path, const char *subdir_name,
+        int *start_index)
+{
+    int last_index;
+    return sf_file_writer_get_binlog_indexes(data_path,
+            subdir_name, start_index, &last_index);
+}
+
+static inline int sf_file_writer_get_binlog_last_index(
+        const char *data_path, const char *subdir_name,
+        int *last_index)
+{
+    int start_index;
+    return sf_file_writer_get_binlog_indexes(data_path,
+            subdir_name, &start_index, last_index);
+}
+
+int sf_file_writer_set_indexes(SFFileWriterInfo *writer,
+        const int start_index, const int last_index);
+
+int sf_file_writer_get_indexes(SFFileWriterInfo *writer,
+        int *start_index, int *last_index);
+
+static inline int sf_file_writer_get_start_index(SFFileWriterInfo *writer)
+{
+    int start_index;
+    int last_index;
+
+    sf_file_writer_get_indexes(writer, &start_index, &last_index);
+    return start_index;
+}
+
+static inline int sf_file_writer_get_last_index(SFFileWriterInfo *writer)
+{
+    int start_index;
+    int last_index;
+
+    sf_file_writer_get_indexes(writer, &start_index, &last_index);
+    return last_index;
+}
+
+#define sf_file_writer_get_current_write_index(writer) \
+    sf_file_writer_get_last_index(writer)
+
+static inline void sf_file_writer_get_current_position(
+        SFFileWriterInfo *writer, SFBinlogFilePosition *position)
+{
+    position->index = writer->binlog.last_index;
+    position->offset = writer->file.size;
+}
+
+static inline const char *sf_file_writer_get_filepath(
+        const char *data_path, const char *subdir_name,
+        char *filepath, const int size)
+{
+    fc_get_full_filepath_ex(data_path, strlen(data_path),
+            subdir_name, strlen(subdir_name), filepath, size);
+    return filepath;
+}
+
+static inline const char *sf_file_writer_get_filename_ex(
+        const char *data_path, const char *subdir_name,
+        const char *file_prefix, const int binlog_index,
+        char *filename, const int size)
+{
+    char *p;
+    int data_path_len;
+    int subdir_name_len;
+    int file_prefix_len;
+
+    data_path_len = strlen(data_path);
+    subdir_name_len = strlen(subdir_name);
+    file_prefix_len = strlen(file_prefix);
+    if (data_path_len + subdir_name_len + file_prefix_len +
+            4 + SF_BINLOG_FILE_EXT_LEN >= size)
+    {
+        snprintf(filename, size, "%s/%s/%s"SF_BINLOG_FILE_EXT_FMT,
+                data_path, subdir_name, file_prefix, binlog_index);
+        return filename;
+    }
+
+    p = filename;
+    memcpy(p, data_path, data_path_len);
+    p += data_path_len;
+    *p++ = '/';
+    memcpy(p, subdir_name, subdir_name_len);
+    p += subdir_name_len;
+    *p++ = '/';
+    memcpy(p, file_prefix, file_prefix_len);
+    p += file_prefix_len;
+    *p++ = '.';
+    fc_ltostr_ex(binlog_index, p, SF_BINLOG_FILE_EXT_LEN);
+
+    return filename;
+}
+
+#define sf_file_writer_get_filename(data_path, subdir_name, \
+        binlog_index, filename, size) \
+    sf_file_writer_get_filename_ex(data_path, subdir_name, \
+        SF_BINLOG_FILE_PREFIX_STR, binlog_index, filename, size)
+
+const char *sf_file_writer_get_index_filename(const char *data_path,
+        const char *subdir_name, char *filename, const int size);
+
+int sf_file_writer_set_binlog_start_index(SFFileWriterInfo *writer,
+        const int start_index);
+
+int sf_file_writer_set_binlog_write_index(SFFileWriterInfo *writer,
+        const int last_index);
+
+static inline int sf_file_writer_rotate_file(SFFileWriterInfo *writer)
+{
+    int last_index;
+    last_index = sf_file_writer_get_current_write_index(writer);
+    return sf_file_writer_set_binlog_write_index(writer, last_index + 1);
+}
+
+int sf_file_writer_get_last_lines(const char *data_path,
+        const char *subdir_name, const int current_write_index,
+        char *buff, const int buff_size, int *count, int *length);
+
+static inline int sf_file_writer_get_last_line_ex(const char *data_path,
+        const char *subdir_name, const int current_write_index,
+        char *buff, const int buff_size, int *length)
+{
+    int count = 1;
+    return sf_file_writer_get_last_lines(data_path, subdir_name,
+            current_write_index, buff, buff_size, &count, length);
+}
+
+int sf_file_writer_get_last_line(const char *data_path,
+        const char *subdir_name, char *buff,
+        const int buff_size, int *length);
+
+int sf_file_writer_write_to_binlog_index_file_ex(const char *data_path,
+        const char *subdir_name, const char *file_prefix,
+        const int start_index, const int last_index,
+        const int compress_index);
+
+#define sf_file_writer_write_to_binlog_index_file(data_path,  \
+        subdir_name, start_index, last_index) \
+    sf_file_writer_write_to_binlog_index_file_ex(data_path, subdir_name, \
+            SF_BINLOG_FILE_PREFIX_STR, start_index, last_index, 0)
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif

src/sf_func.c

@@ -38,16 +38,17 @@ int sf_connect_to_server(const char *ip_addr, const int port, int *sock)
     if(*sock < 0) {
         return errno != 0 ? errno : ENOMEM;
     }
-    tcpsetserveropt(*sock, g_sf_global_vars.network_timeout);
+    tcpsetserveropt(*sock, g_sf_global_vars.net_buffer_cfg.network_timeout);
 
     if ((result=tcpsetnonblockopt(*sock)) != 0) {
         close(*sock);
         *sock = -1;
         return result;
     }
+    FC_SET_CLOEXEC(*sock);
 
-    if ((result=connectserverbyip_nb(*sock, ip_addr, port,
-                    g_sf_global_vars.connect_timeout)) != 0)
+    if ((result=connectserverbyip_nb(*sock, ip_addr, port, g_sf_global_vars.
+                    net_buffer_cfg.connect_timeout)) != 0)
     {
         close(*sock);
         *sock = -1;

src/sf_func.h

@@ -18,7 +18,8 @@
 #ifndef _SF_FUNC_H
 #define _SF_FUNC_H
 
-#include "fastcommon/common_define.h"
+#include "fastcommon/pthread_func.h"
+#include "fastcommon/fc_atomic.h"
 #include "sf_types.h"
 #include "sf_global.h"
 
@@ -36,7 +37,7 @@ static inline void sf_terminate_myself_ex(const char *file,
 {
     g_sf_global_vars.continue_flag = false;
     if (kill(getpid(), SIGQUIT) == 0) { //signal myself to quit
-        logInfo("file: "__FILE__", line: %d, "
+        logWarning("file: "__FILE__", line: %d, "
                 "kill myself from caller {file: %s, line: %d, func: %s}",
                 __LINE__, file, line, func);
     } else {
@@ -55,7 +56,8 @@ static inline int sf_binlog_buffer_init(SFBinlogBuffer *buffer, const int size)
         return ENOMEM;
     }
 
-    buffer->current = buffer->end = buffer->buff;
+    buffer->current = buffer->data_end = buffer->buff;
+    buffer->buff_end = buffer->buff + size;
     buffer->size = size;
     return 0;
 }
@@ -64,11 +66,87 @@ static inline void sf_binlog_buffer_destroy(SFBinlogBuffer *buffer)
 {
     if (buffer->buff != NULL) {
         free(buffer->buff);
-        buffer->current = buffer->end = buffer->buff = NULL;
+        buffer->current = buffer->buff = NULL;
+        buffer->data_end = buffer->buff_end = NULL;
         buffer->size = 0;
     }
 }
 
+static inline int sf_synchronize_ctx_init(SFSynchronizeContext *sctx)
+{
+    sctx->waiting_count = 0;
+    return init_pthread_lock_cond_pair(&sctx->lcp);
+}
+
+static inline void sf_synchronize_ctx_destroy(SFSynchronizeContext *sctx)
+{
+    destroy_pthread_lock_cond_pair(&sctx->lcp);
+}
+
+static inline void sf_synchronize_counter_add(
+        SFSynchronizeContext *sctx, const int count)
+{
+    PTHREAD_MUTEX_LOCK(&sctx->lcp.lock);
+    sctx->waiting_count += count;
+    PTHREAD_MUTEX_UNLOCK(&sctx->lcp.lock);
+}
+
+static inline void sf_synchronize_counter_sub(
+        SFSynchronizeContext *sctx, const int count)
+{
+    PTHREAD_MUTEX_LOCK(&sctx->lcp.lock);
+    sctx->waiting_count -= count;
+    PTHREAD_MUTEX_UNLOCK(&sctx->lcp.lock);
+}
+
+static inline void sf_synchronize_counter_notify(
+        SFSynchronizeContext *sctx, const int count)
+{
+    PTHREAD_MUTEX_LOCK(&sctx->lcp.lock);
+    sctx->waiting_count -= count;
+    if (sctx->waiting_count == 0) {
+        pthread_cond_signal(&sctx->lcp.cond);
+    }
+    PTHREAD_MUTEX_UNLOCK(&sctx->lcp.lock);
+}
+
+static inline void sf_synchronize_counter_wait(SFSynchronizeContext *sctx)
+{
+    PTHREAD_MUTEX_LOCK(&sctx->lcp.lock);
+    while (sctx->waiting_count != 0 && SF_G_CONTINUE_FLAG) {
+        pthread_cond_wait(&sctx->lcp.cond, &sctx->lcp.lock);
+    }
+    PTHREAD_MUTEX_UNLOCK(&sctx->lcp.lock);
+}
+
+#define sf_synchronize_finished_notify_no_lock(sctx, err_no) \
+    (sctx)->finished = true;  \
+    (sctx)->result = err_no;  \
+    pthread_cond_signal(&(sctx)->lcp.cond)
+
+static inline void sf_synchronize_finished_notify(
+        SFSynchronizeContext *sctx, const int result)
+{
+    PTHREAD_MUTEX_LOCK(&sctx->lcp.lock);
+    sf_synchronize_finished_notify_no_lock(sctx, result);
+    PTHREAD_MUTEX_UNLOCK(&sctx->lcp.lock);
+}
+
+static inline int sf_synchronize_finished_wait(SFSynchronizeContext *sctx)
+{
+    int result;
+
+    PTHREAD_MUTEX_LOCK(&sctx->lcp.lock);
+    while (!sctx->finished && SF_G_CONTINUE_FLAG) {
+        pthread_cond_wait(&sctx->lcp.cond, &sctx->lcp.lock);
+    }
+    result = sctx->result;
+    sctx->finished = false;  //for next notify
+    PTHREAD_MUTEX_UNLOCK(&sctx->lcp.lock);
+
+    return result;
+}
+
 #ifdef __cplusplus
 }
 #endif

src/sf_global.h

@@ -30,30 +30,36 @@ typedef struct sf_connection_stat {
 } SFConnectionStat;
 
 typedef struct sf_global_variables {
-    int connect_timeout;
-    int network_timeout;
-    char base_path[MAX_PATH_SIZE];
+    struct {
+        char str[MAX_PATH_SIZE];
+        int len;
+        bool inited;
+        bool created;
+    } base_path;
 
     volatile bool continue_flag;
     bool tcp_quick_ack;
-    int max_connections;
-    int max_pkg_size;
-    int min_buff_size;
-    int max_buff_size;
+    bool epoll_edge_trigger;
+
+    SFNetBufferConfig net_buffer_cfg;
+
     int task_buffer_extra_size;
     int thread_stack_size;
-    int sync_log_buff_interval; //sync log buff to disk every interval seconds
 
     time_t up_time;
-    gid_t run_by_gid;
-    uid_t run_by_uid;
-    char run_by_group[32];
-    char run_by_user[32];
-
-    bool rotate_error_log;
-    int log_file_keep_days;
+    struct {
+        bool inited;
+        gid_t gid;
+        uid_t uid;
+        char group[32];
+        char user[32];
+    } run_by;
 
+    SFLogConfig error_log;
     SFConnectionStat connection_stat;
+    sf_error_handler_callback error_handler;
+    string_t empty;
+    volatile time_t last_binlog_writer_log_timestamp;
 } SFGlobalVariables;
 
 typedef struct sf_context_ini_config {
@@ -61,6 +67,9 @@ typedef struct sf_context_ini_config {
     int default_inner_port;
     int default_outer_port;
     int default_work_threads;
+    int max_pkg_size_min_value;
+    FCCommunicationType comm_type;
+    const char *max_pkg_size_item_name;
 } SFContextIniConfig;
 
 #ifdef __cplusplus
@@ -70,97 +79,221 @@ extern "C" {
 extern SFGlobalVariables         g_sf_global_vars;
 extern SFContext                 g_sf_context;
 
-#define SF_G_BASE_PATH           g_sf_global_vars.base_path
+#define SF_G_BASE_PATH_STR       g_sf_global_vars.base_path.str
+#define SF_G_BASE_PATH_LEN       g_sf_global_vars.base_path.len
+#define SF_G_BASE_PATH_INITED    g_sf_global_vars.base_path.inited
+#define SF_G_BASE_PATH_CREATED   g_sf_global_vars.base_path.created
 #define SF_G_CONTINUE_FLAG       g_sf_global_vars.continue_flag
-#define SF_G_CONNECT_TIMEOUT     g_sf_global_vars.connect_timeout
-#define SF_G_NETWORK_TIMEOUT     g_sf_global_vars.network_timeout
-#define SF_G_MAX_CONNECTIONS     g_sf_global_vars.max_connections
+#define SF_G_CONNECT_TIMEOUT     g_sf_global_vars.net_buffer_cfg.connect_timeout
+#define SF_G_NETWORK_TIMEOUT     g_sf_global_vars.net_buffer_cfg.network_timeout
+#define SF_G_MAX_CONNECTIONS     g_sf_global_vars.net_buffer_cfg.max_connections
 #define SF_G_THREAD_STACK_SIZE   g_sf_global_vars.thread_stack_size
+#define SF_G_UP_TIME             g_sf_global_vars.up_time
 
+#define SF_G_SOCK_HANDLER        (g_sf_context.handlers  \
+        [SF_IPV4_ADDRESS_FAMILY_INDEX].handlers + \
+        SF_SOCKET_NETWORK_HANDLER_INDEX)
+#define SF_G_OUTER_PORT          SF_G_SOCK_HANDLER->outer.port
+#define SF_G_INNER_PORT          SF_G_SOCK_HANDLER->inner.port
+#define SF_G_OUTER_BIND_ADDR4     g_sf_context.handlers  \
+        [SF_IPV4_ADDRESS_FAMILY_INDEX].outer_bind_addr
+#define SF_G_INNER_BIND_ADDR4     g_sf_context.handlers  \
+        [SF_IPV4_ADDRESS_FAMILY_INDEX].inner_bind_addr
+#define SF_G_OUTER_BIND_ADDR6     g_sf_context.handlers  \
+        [SF_IPV6_ADDRESS_FAMILY_INDEX].outer_bind_addr
+#define SF_G_INNER_BIND_ADDR6     g_sf_context.handlers  \
+        [SF_IPV6_ADDRESS_FAMILY_INDEX].inner_bind_addr
+
+#define SF_G_IPV4_ENABLED     (g_sf_context.handlers  \
+        [SF_IPV4_ADDRESS_FAMILY_INDEX].af == AF_INET)
+#define SF_G_IPV6_ENABLED     (g_sf_context.handlers  \
+        [SF_IPV6_ADDRESS_FAMILY_INDEX].af == AF_INET6)
+
+#define SF_G_ACCEPT_THREADS      g_sf_context.accept_threads
 #define SF_G_WORK_THREADS        g_sf_context.work_threads
 #define SF_G_ALIVE_THREAD_COUNT  g_sf_context.thread_count
 #define SF_G_THREAD_INDEX(tdata) (int)(tdata - g_sf_context.thread_data)
 #define SF_G_CONN_CURRENT_COUNT  g_sf_global_vars.connection_stat.current_count
 #define SF_G_CONN_MAX_COUNT      g_sf_global_vars.connection_stat.max_count
 
-#define SF_WORK_THREADS(sf_context)        sf_context.work_threads
-#define SF_ALIVE_THREAD_COUNT(sf_context)  sf_context.thread_count
-#define SF_THREAD_INDEX(sf_context, tdata) (int)(tdata - sf_context.thread_data)
+#define SF_G_ERROR_HANDLER       g_sf_global_vars.error_handler
+#define SF_G_EMPTY_STRING        g_sf_global_vars.empty
+#define LAST_BINLOG_WRITER_LOG_TIMESTAMP  g_sf_global_vars. \
+    last_binlog_writer_log_timestamp
+
+#define SF_G_EPOLL_EDGE_TRIGGER  g_sf_global_vars.epoll_edge_trigger
+
+#define SF_WORK_THREADS(sf_context)        (sf_context).work_threads
+#define SF_ALIVE_THREAD_COUNT(sf_context)  (sf_context).thread_count
+#define SF_THREAD_INDEX(sf_context, tdata) (int)(tdata - (sf_context).thread_data)
+
+#define SF_IPV4_ENABLED(sf_context)   ((sf_context).handlers  \
+        [SF_IPV4_ADDRESS_FAMILY_INDEX].af == AF_INET)
+#define SF_IPV6_ENABLED(sf_context)   ((sf_context).handlers  \
+        [SF_IPV6_ADDRESS_FAMILY_INDEX].af == AF_INET6)
 
 #define SF_CHOWN_RETURN_ON_ERROR(path, current_uid, current_gid) \
     do { \
-    if (!(g_sf_global_vars.run_by_gid == current_gid && \
-                g_sf_global_vars.run_by_uid == current_uid)) \
-    { \
-        if (chown(path, g_sf_global_vars.run_by_uid, \
-                    g_sf_global_vars.run_by_gid) != 0) \
+        if (g_sf_global_vars.run_by.inited && !(g_sf_global_vars.  \
+                    run_by.gid == current_gid && g_sf_global_vars. \
+                    run_by.uid == current_uid)) \
         { \
-            logError("file: "__FILE__", line: %d, " \
-                "chown \"%s\" fail, " \
-                "errno: %d, error info: %s", \
-                __LINE__, path, errno, STRERROR(errno)); \
-            return errno != 0 ? errno : EPERM; \
+            if (chown(path, g_sf_global_vars.run_by.uid, \
+                        g_sf_global_vars.run_by.gid) != 0) \
+            { \
+                logError("file: "__FILE__", line: %d, " \
+                        "chown \"%s\" fail, " \
+                        "errno: %d, error info: %s", \
+                        __LINE__, path, errno, STRERROR(errno)); \
+                return errno != 0 ? errno : EPERM; \
+            } \
         } \
-    } \
     } while (0)
 
-#define SF_SET_CONTEXT_INI_CONFIG(config, filename, pIniContext, \
-        section_name, def_inner_port, def_outer_port, def_work_threads) \
+#define SF_CHOWN_TO_RUNBY_RETURN_ON_ERROR(path) \
+    SF_CHOWN_RETURN_ON_ERROR(path, geteuid(), getegid())
+
+
+#define SF_FCHOWN_RETURN_ON_ERROR(fd, path, current_uid, current_gid) \
+    do { \
+        if (g_sf_global_vars.run_by.inited && !(g_sf_global_vars.  \
+                    run_by.gid == current_gid && g_sf_global_vars. \
+                    run_by.uid == current_uid)) \
+        { \
+            if (fchown(fd, g_sf_global_vars.run_by.uid, \
+                        g_sf_global_vars.run_by.gid) != 0) \
+            { \
+                logError("file: "__FILE__", line: %d, " \
+                        "fchown \"%s\" fail, " \
+                        "errno: %d, error info: %s", \
+                        __LINE__, path, errno, STRERROR(errno)); \
+                return errno != 0 ? errno : EPERM; \
+            } \
+        } \
+    } while (0)
+
+#define SF_FCHOWN_TO_RUNBY_RETURN_ON_ERROR(fd, path) \
+    SF_FCHOWN_RETURN_ON_ERROR(fd, path, geteuid(), getegid())
+
+#define SF_SET_CONTEXT_INI_CONFIG_EX(config, the_comm_type, filename, \
+        pIniContext, section_name, def_inner_port, def_outer_port,    \
+        def_work_threads, max_pkg_size_item_nm, max_pkg_size_min_val) \
     do { \
         FAST_INI_SET_FULL_CTX_EX(config.ini_ctx, filename, \
                 section_name, pIniContext);   \
+        config.comm_type = the_comm_type; \
         config.default_inner_port = def_inner_port; \
         config.default_outer_port = def_outer_port; \
         config.default_work_threads = def_work_threads; \
+        config.max_pkg_size_item_name = max_pkg_size_item_nm; \
+        config.max_pkg_size_min_value = max_pkg_size_min_val; \
     } while (0)
 
-int sf_load_global_config_ex(const char *server_name,
-        IniFullContext *ini_ctx, const bool load_network_params,
-        const int task_buffer_extra_size);
+#define SF_SET_CONTEXT_INI_CONFIG(config, the_comm_type,     \
+        filename, pIniContext, section_name, def_inner_port, \
+        def_outer_port, def_work_threads) \
+     SF_SET_CONTEXT_INI_CONFIG_EX(config, the_comm_type, filename, \
+             pIniContext, section_name, def_inner_port, def_outer_port, \
+             def_work_threads, "max_pkg_size", 0)
 
-static inline int sf_load_global_config(const char *server_name,
+int sf_load_global_config_ex(const char *log_filename_prefix,
+        IniFullContext *ini_ctx, const bool load_network_params,
+        const char *max_pkg_size_item_nm, const int fixed_buff_size,
+        const int task_buffer_extra_size, const bool need_set_run_by);
+
+static inline int sf_load_global_config(const char *log_filename_prefix,
         IniFullContext *ini_ctx)
 {
     const bool load_network_params = true;
+    const char *max_pkg_size_item_nm = "max_pkg_size";
+    const int fixed_buff_size = 0;
     const int task_buffer_extra_size = 0;
+    const bool need_set_run_by = true;
 
-    return sf_load_global_config_ex(server_name, ini_ctx,
-            load_network_params, task_buffer_extra_size);
+    return sf_load_global_config_ex(log_filename_prefix, ini_ctx,
+            load_network_params, max_pkg_size_item_nm, fixed_buff_size,
+            task_buffer_extra_size, need_set_run_by);
 }
 
-int sf_load_config_ex(const char *server_name,
-        SFContextIniConfig *config, const int task_buffer_extra_size);
+int sf_load_config_ex(const char *log_filename_prefix,
+        SFContextIniConfig *config, const int fixed_buff_size,
+        const int task_buffer_extra_size, const bool need_set_run_by);
 
-static inline int sf_load_config(const char *server_name,
+static inline int sf_load_config(const char *log_filename_prefix,
+        const FCCommunicationType comm_type,
         const char *filename, IniContext *pIniContext,
         const char *section_name, const int default_inner_port,
-        const int default_outer_port, const int task_buffer_extra_size)
+        const int default_outer_port, const int fixed_buff_size,
+        const int task_buffer_extra_size)
 {
+    const bool need_set_run_by = true;
     SFContextIniConfig config;
 
-    SF_SET_CONTEXT_INI_CONFIG(config, filename, pIniContext,
+    SF_SET_CONTEXT_INI_CONFIG(config, comm_type, filename, pIniContext,
             section_name, default_inner_port, default_outer_port,
             DEFAULT_WORK_THREADS);
-    return sf_load_config_ex(server_name, &config, task_buffer_extra_size);
+    return sf_load_config_ex(log_filename_prefix, &config, fixed_buff_size,
+            task_buffer_extra_size, need_set_run_by);
 }
 
 int sf_load_context_from_config_ex(SFContext *sf_context,
-        SFContextIniConfig *config);
+        SFContextIniConfig *config, const int fixed_buff_size,
+        const int task_buffer_extra_size);
 
 static inline int sf_load_context_from_config(SFContext *sf_context,
+        const FCCommunicationType comm_type,
         const char *filename, IniContext *pIniContext,
         const char *section_name, const int default_inner_port,
-        const int default_outer_port)
+        const int default_outer_port, const int fixed_buff_size,
+        const int task_buffer_extra_size)
 {
     SFContextIniConfig config;
 
-    SF_SET_CONTEXT_INI_CONFIG(config, filename, pIniContext,
+    SF_SET_CONTEXT_INI_CONFIG(config, comm_type, filename, pIniContext,
             section_name, default_inner_port, default_outer_port,
             DEFAULT_WORK_THREADS);
-    return sf_load_context_from_config_ex(sf_context, &config);
+    return sf_load_context_from_config_ex(sf_context, &config,
+            fixed_buff_size, task_buffer_extra_size);
 }
 
-void sf_global_config_to_string(char *output, const int size);
+int sf_alloc_rdma_pd(SFContext *sf_context,
+        FCAddressPtrArray *address_array);
+
+void sf_set_address_family_by_ip(SFContext *sf_context,
+        FCAddressPtrArray *address_array);
+
+int sf_load_log_config(IniFullContext *ini_ctx, LogContext *log_ctx,
+        SFLogConfig *log_cfg);
+
+int sf_load_slow_log_config_ex(IniFullContext *ini_ctx, LogContext *log_ctx,
+        SFSlowLogConfig *slow_log_cfg);
+
+static inline int sf_load_slow_log_config(const char *config_file,
+        IniContext *ini_context, LogContext *log_ctx,
+        SFSlowLogConfig *slow_log_cfg)
+{
+    IniFullContext ini_ctx;
+
+    FAST_INI_SET_FULL_CTX_EX(ini_ctx, config_file, "slow-log", ini_context);
+    return sf_load_slow_log_config_ex(&ini_ctx, log_ctx, slow_log_cfg);
+}
+
+void sf_set_log_rotate_size(LogContext *context, const int64_t log_rotate_size);
+
+void sf_log_config_to_string_ex(SFLogConfig *log_cfg, const char *caption,
+        const char *other_config, char *output, const int size);
+
+void sf_slow_log_config_to_string(SFSlowLogConfig *slow_log_cfg,
+        const char *caption, char *output, const int size);
+
+void sf_global_config_to_string_ex(const char *max_pkg_size_item_nm,
+        char *output, const int size);
+
+static inline void sf_global_config_to_string(char *output, const int size)
+{
+    const char *max_pkg_size_item_nm = "max_pkg_size";
+    sf_global_config_to_string_ex(max_pkg_size_item_nm, output, size);
+}
 
 void sf_context_config_to_string(const SFContext *sf_context,
         char *output, const int size);
@@ -169,6 +302,34 @@ void sf_log_config_ex(const char *other_config);
 
 #define sf_log_config() sf_log_config_ex(NULL)
 
+#define sf_log_config_to_string(log_cfg, caption, output, size)  \
+    sf_log_config_to_string_ex(log_cfg, caption, NULL, output, size)
+
+int sf_get_base_path_from_conf_file(const char *config_filename);
+int sf_load_global_base_path(IniFullContext *ini_ctx);
+
+int sf_load_data_path_config_ex(IniFullContext *ini_ctx,
+        const char *item_name, const char *default_value, string_t *path);
+
+#define sf_load_data_path_config(ini_ctx, path) \
+    sf_load_data_path_config_ex(ini_ctx, "data_path", "data", path)
+
+static inline void sf_set_global_base_path(const char *base_path)
+{
+    string_t path_string;
+
+    FC_SET_STRING(path_string, (char *)base_path);
+    SF_G_BASE_PATH_LEN = normalize_path(NULL, &path_string,
+            SF_G_BASE_PATH_STR, sizeof(SF_G_BASE_PATH_STR));
+    SF_G_BASE_PATH_INITED = true;
+}
+
+static inline void sf_set_error_handler(
+        sf_error_handler_callback error_handler)
+{
+    SF_G_ERROR_HANDLER = error_handler;
+}
+
 #ifdef __cplusplus
 }
 #endif

src/sf_iov.c

@@ -0,0 +1,197 @@
+/*
+ * Copyright (c) 2020 YuQing <384681@qq.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <https://www.gnu.org/licenses/>.
+ */
+
+#include "fastcommon/logger.h"
+#include "fastcommon/sockopt.h"
+#include "sf_define.h"
+#include "sf_iov.h"
+
+int sf_iova_consume(SFDynamicIOVArray *iova, const int consume_len)
+{
+    struct iovec *iob;
+    struct iovec *end;
+    int sum_bytes;
+    int remain_len;
+    int result;
+
+    if (iova->cnt <= 0) {
+        logError("file: "__FILE__", line: %d, "
+                "invalid iov count: %d", __LINE__, iova->cnt);
+        return EINVAL;
+    }
+
+    if ((result=sf_iova_check_alloc(iova)) != 0) {
+        return result;
+    }
+
+    end = iova->iov + iova->cnt;
+    iob = iova->iov;
+    sum_bytes = iob->iov_len;
+    for (iob=iob + 1; sum_bytes <= consume_len && iob < end; iob++) {
+        sum_bytes += iob->iov_len;
+    }
+
+    if (sum_bytes < consume_len) {
+        logError("file: "__FILE__", line: %d, "
+                "iov length: %d < consume length: %d",
+                __LINE__, sum_bytes, consume_len);
+        return EOVERFLOW;
+    }
+
+    iova->cnt -= (iob - iova->iov);
+    iova->iov = iob;
+    if (iova->cnt == 0) {
+        struct iovec *last;
+
+        /* update the last iov for next slice */
+        last = iob - 1;
+        last->iov_base = (char *)last->iov_base + last->iov_len;
+        last->iov_len = 0;
+    } else {
+        /* adjust the first element */
+        remain_len = sum_bytes - consume_len;
+        if (remain_len < iob->iov_len) {
+            iob->iov_base = (char *)iob->iov_base +
+                (iob->iov_len - remain_len);
+            iob->iov_len = remain_len;
+        }
+    }
+
+    return 0;
+}
+
+static inline int iova_slice(SFDynamicIOVArray *iova, const int slice_len)
+{
+    struct iovec *iob;
+    struct iovec *end;
+    int sum_bytes;
+    int exceed_len;
+
+    sum_bytes = 0;
+    end = iova->ptr + iova->input.cnt;
+    for (iob=iova->iov; iob<end; iob++) {
+        sum_bytes += iob->iov_len;
+        if (sum_bytes > slice_len) {
+            exceed_len = sum_bytes - slice_len;
+            iob->iov_len -= exceed_len;
+            break;
+        } else if (sum_bytes == slice_len) {
+            break;
+        }
+    }
+
+    if (iob < end) {
+        iova->cnt = (iob - iova->iov) + 1;
+        return 0;
+    } else {
+        logError("file: "__FILE__", line: %d, "
+                "iov remain bytes: %d < slice length: %d",
+                __LINE__, sum_bytes, slice_len);
+        iova->cnt = 0;
+        return EOVERFLOW;
+    }
+}
+
+int sf_iova_first_slice(SFDynamicIOVArray *iova, const int slice_len)
+{
+    int result;
+
+    if ((result=sf_iova_check_alloc(iova)) != 0) {
+        return result;
+    }
+
+    return iova_slice(iova, slice_len);
+}
+
+int sf_iova_next_slice(SFDynamicIOVArray *iova,
+        const int consume_len, const int slice_len)
+{
+    struct iovec *last;
+    const struct iovec *origin;
+    int remain_len;
+    int result;
+
+    if ((result=sf_iova_consume(iova, consume_len)) != 0) {
+        return result;
+    }
+
+    last = iova->iov + iova->cnt - 1;
+    origin = iova->input.iov + (last - iova->ptr);
+    remain_len = ((char *)origin->iov_base + origin->iov_len) -
+        (char *)last->iov_base;
+    if (last->iov_len != remain_len) {
+        last->iov_len = remain_len;
+        if (iova->cnt == 0) {
+            iova->iov = last;
+        }
+    }
+
+    return iova_slice(iova, slice_len);
+}
+
+int sf_iova_memset_ex(const struct iovec *iov, const int iovcnt,
+        int c, const int offset, const int length)
+{
+    const struct iovec *iob;
+    const struct iovec *end;
+    int sum_bytes;
+    int remain_len;
+    int left_bytes;
+    char *start;
+
+    if (length == 0) {
+        return 0;
+    }
+
+    sum_bytes = 0;
+    end = iov + iovcnt;
+    for (iob=iov; iob<end; iob++) {
+        sum_bytes += iob->iov_len;
+        if (sum_bytes > offset) {
+            break;
+        }
+    }
+
+    if (iob == end) {
+        logError("file: "__FILE__", line: %d, "
+                "iov length: %d < (offset: %d + length: %d)",
+                __LINE__, sum_bytes, offset, length);
+        return EOVERFLOW;
+    }
+
+    remain_len = sum_bytes - offset;
+    start = (char *)iob->iov_base + (iob->iov_len - remain_len);
+    if (length <= remain_len) {
+        memset(start, c, length);
+        return 0;
+    }
+
+    memset(start, c, remain_len);
+    left_bytes = length - remain_len;
+    while (++iob < end) {
+        if (left_bytes <= iob->iov_len) {
+            memset(iob->iov_base, c, left_bytes);
+            return 0;
+        }
+
+        memset(iob->iov_base, c, iob->iov_len);
+        left_bytes -= iob->iov_len;
+    }
+
+    logError("file: "__FILE__", line: %d, "
+            "iov length is too short, overflow bytes: %d",
+            __LINE__, left_bytes);
+    return EOVERFLOW;
+}

src/sf_iov.h

@@ -0,0 +1,116 @@
+/*
+ * Copyright (c) 2020 YuQing <384681@qq.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <https://www.gnu.org/licenses/>.
+ */
+
+#ifndef _SF_IOV_H
+#define _SF_IOV_H
+
+#include "fastcommon/shared_func.h"
+#include "sf_types.h"
+
+#define SF_IOV_FIXED_SIZE    256
+
+typedef struct sf_dynamic_iov_array {
+    struct iovec holder[SF_IOV_FIXED_SIZE];
+    struct iovec *ptr;
+
+    struct {
+        const struct iovec *iov;
+        int cnt;
+    } input;
+
+    struct iovec *iov;
+    int cnt;
+} SFDynamicIOVArray;
+
+#define sf_iova_init(iova, _iov, _cnt)  \
+    (iova).input.iov = _iov;            \
+    (iova).iov = (struct iovec *)_iov;  \
+    (iova).cnt = (iova).input.cnt = _cnt
+
+#define sf_iova_destroy(iova) \
+    if ((iova).iov != (struct iovec *)(iova).input.iov && \
+            (iova).ptr != (iova).holder)  \
+            free((iova).ptr)
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+static inline int sf_iova_check_alloc(SFDynamicIOVArray *iova)
+{
+    if (iova->iov == (struct iovec *)iova->input.iov) {
+        if (iova->input.cnt <= SF_IOV_FIXED_SIZE) {
+            iova->ptr = iova->holder;
+        } else {
+            iova->ptr = fc_malloc(iova->input.cnt *
+                    sizeof(struct iovec));
+            if (iova->ptr == NULL) {
+                return ENOMEM;
+            }
+        }
+
+        memcpy(iova->ptr, iova->input.iov, iova->input.cnt *
+                sizeof(struct iovec));
+        iova->iov = iova->ptr;
+    }
+
+    return 0;
+}
+
+int sf_iova_consume(SFDynamicIOVArray *iova, const int consume_len);
+
+int sf_iova_first_slice(SFDynamicIOVArray *iova, const int slice_len);
+
+int sf_iova_next_slice(SFDynamicIOVArray *iova,
+        const int consume_len, const int slice_len);
+
+int sf_iova_memset_ex(const struct iovec *iov, const int iovcnt,
+        int c, const int offset, const int length);
+
+#define sf_iova_memset(iova, c, offset, length) \
+    sf_iova_memset_ex((iova).iov, (iova).cnt, c, offset, length)
+
+static inline void sf_iova_memcpy_ex(const struct iovec *iov,
+        const int iovcnt, const char *buff, const int length)
+{
+    const struct iovec *iob;
+    const struct iovec *end;
+    const char *current;
+    int remain;
+    int bytes;
+
+    current = buff;
+    remain = length;
+    end = iov + iovcnt;
+    for (iob=iov; iob<end; iob++) {
+        bytes = FC_MIN(remain, iob->iov_len);
+        memcpy(iob->iov_base, current, bytes);
+
+        remain -= bytes;
+        if (remain == 0) {
+            break;
+        }
+        current += bytes;
+    }
+}
+
+#define sf_iova_memcpy(iova, buff, length) \
+    sf_iova_memcpy_ex((iova).iov, (iova).cnt, buff, length)
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif

src/sf_nio.h

@@ -22,8 +22,13 @@
 #include <stdlib.h>
 #include <string.h>
 #include "fastcommon/fast_task_queue.h"
+#include "fastcommon/ioevent_loop.h"
 #include "sf_define.h"
 #include "sf_types.h"
+#include "sf_global.h"
+
+#define SF_CTX              (task->handler->fh->ctx)
+#define SF_NET_BUFFER_CFG   SF_CTX->net_buffer_cfg
 
 #ifdef __cplusplus
 extern "C" {
@@ -31,40 +36,63 @@ extern "C" {
 
 void sf_set_parameters_ex(SFContext *sf_context, const int header_size,
         sf_set_body_length_callback set_body_length_func,
-        sf_deal_task_func deal_func, TaskCleanUpCallback cleanup_func,
-        sf_recv_timeout_callback timeout_callback);
+        sf_alloc_recv_buffer_callback alloc_recv_buffer_func,
+        sf_send_done_callback send_done_callback,
+        sf_deal_task_callback deal_func, TaskCleanUpCallback cleanup_func,
+        sf_recv_timeout_callback timeout_callback, sf_release_buffer_callback
+        release_buffer_callback);
 
 #define sf_set_parameters(header_size, set_body_length_func, \
-        deal_func, cleanup_func, timeout_callback)   \
+        alloc_recv_buffer_func, deal_func, cleanup_func, timeout_callback) \
     sf_set_parameters_ex(&g_sf_context, header_size, \
-            set_body_length_func, deal_func, \
-            cleanup_func, timeout_callback)
+            set_body_length_func, alloc_recv_buffer_func, \
+            deal_func, cleanup_func, timeout_callback, NULL)
 
-static inline void sf_set_remove_from_ready_list_ex(SFContext *sf_context,
-        const bool enabled)
+static inline void sf_set_deal_task_callback_ex(SFContext *sf_context,
+        sf_deal_task_callback deal_func)
 {
-    sf_context->remove_from_ready_list = enabled;
+    sf_context->callbacks.deal_task = deal_func;
 }
 
-#define sf_set_remove_from_ready_list(enabled) \
-    sf_set_remove_from_ready_list_ex(&g_sf_context, enabled);
+#define sf_set_deal_task_callback(deal_func) \
+    sf_set_deal_task_callback_ex(&g_sf_context, deal_func)
 
-static inline TaskCleanUpCallback sf_get_task_cleanup_func_ex(
+
+static inline void sf_set_connect_done_callback_ex(SFContext *sf_context,
+        sf_connect_done_callback done_callback)
+{
+    sf_context->callbacks.connect_done = done_callback;
+}
+
+#define sf_set_connect_done_callback(done_callback) \
+    sf_set_connect_done_callback_ex(&g_sf_context, done_callback)
+
+
+static inline TaskCleanUpCallback sf_get_task_cleanup_callback_ex(
         SFContext *sf_context)
 {
-    return sf_context->task_cleanup_func;
+    return sf_context->callbacks.task_cleanup;
 }
 
-#define sf_get_task_cleanup_func() \
-    sf_get_task_cleanup_func_ex(&g_sf_context)
+#define sf_get_task_cleanup_callback() \
+    sf_get_task_cleanup_callback_ex(&g_sf_context)
 
-#define sf_nio_task_is_idle(task) \
-    (task->offset == 0 && task->length == 0)
+#define sf_nio_task_send_done(task) \
+    (task->send.ptr->offset == 0 && task->send.ptr->length == 0)
 
-void sf_recv_notify_read(int sock, short event, void *arg);
+static inline void sf_nio_reset_task_length(struct fast_task_info *task)
+{
+    task->send.ptr->length = 0;
+    task->send.ptr->offset = 0;
+    if (task->recv.ptr != task->send.ptr) {
+        task->recv.ptr->length = 0;
+        task->recv.ptr->offset = 0;
+    }
+}
+
+void sf_socket_close_connection(struct fast_task_info *task);
+void sf_recv_notify_read(int sock, const int event, void *arg);
 int sf_send_add_event(struct fast_task_info *task);
-int sf_client_sock_write(int sock, short event, void *arg);
-int sf_client_sock_read(int sock, short event, void *arg);
 
 void sf_task_finish_clean_up(struct fast_task_info *task);
 
@@ -77,6 +105,42 @@ void sf_task_switch_thread(struct fast_task_info *task,
 
 void sf_task_detach_thread(struct fast_task_info *task);
 
+static inline int sf_set_body_length(struct fast_task_info *task)
+{
+    if (SF_CTX->callbacks.set_body_length(task) != 0) {
+        return -1;
+    }
+    if (task->recv.ptr->length < 0) {
+        logError("file: "__FILE__", line: %d, "
+                "client ip: %s, pkg length: %d < 0",
+                __LINE__, task->client_ip,
+                task->recv.ptr->length);
+        return -1;
+    }
+
+    task->recv.ptr->length += SF_CTX->header_size;
+    if (task->recv.ptr->length > SF_NET_BUFFER_CFG.max_pkg_size) {
+        logError("file: "__FILE__", line: %d, "
+                "client ip: %s, pkg length: %d > "
+                "max pkg size: %d", __LINE__,
+                task->client_ip, task->recv.ptr->length,
+                SF_NET_BUFFER_CFG.max_pkg_size);
+        return -1;
+    }
+
+    return 0;
+}
+
+int sf_socket_async_connect_server(struct fast_task_info *task);
+int sf_socket_async_connect_check(struct fast_task_info *task);
+
+ssize_t sf_socket_send_data(struct fast_task_info *task,
+        SFCommAction *action, bool *send_done);
+ssize_t sf_socket_recv_data(struct fast_task_info *task,
+        const bool call_post_recv, SFCommAction *action);
+
+int sf_rdma_busy_polling_callback(struct nio_thread_data *thread_data);
+
 static inline int sf_nio_forward_request(struct fast_task_info *task,
         const int new_thread_index)
 {
@@ -84,11 +148,18 @@ static inline int sf_nio_forward_request(struct fast_task_info *task,
     return sf_nio_notify(task, SF_NIO_STAGE_FORWARDED);
 }
 
-static inline bool sf_client_sock_in_read_stage(struct fast_task_info *task)
+static inline void sf_nio_add_to_deleted_list(struct nio_thread_data
+        *thread_data, struct fast_task_info *task)
 {
-    return (task->event.callback == (IOEventCallback)sf_client_sock_read);
+    if (task->thread_data == thread_data) {
+        ioevent_add_to_deleted_list(task);
+    } else {
+        sf_nio_notify(task, SF_NIO_STAGE_CLOSE);
+    }
 }
 
+bool sf_client_sock_in_read_stage(struct fast_task_info *task);
+
 #ifdef __cplusplus
 }
 #endif

src/sf_ordered_writer.c

@@ -0,0 +1,260 @@
+/*
+ * Copyright (c) 2020 YuQing <384681@qq.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <https://www.gnu.org/licenses/>.
+ */
+
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <sys/socket.h>
+#include <netinet/in.h>
+#include <arpa/inet.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <errno.h>
+#include <limits.h>
+#include <fcntl.h>
+#include <pthread.h>
+#include "fastcommon/logger.h"
+#include "fastcommon/sockopt.h"
+#include "fastcommon/shared_func.h"
+#include "fastcommon/pthread_func.h"
+#include "fastcommon/sched_thread.h"
+#include "sf_global.h"
+#include "sf_func.h"
+#include "sf_ordered_writer.h"
+
+#define deal_binlog_one_record(writer, wb) \
+    sf_file_writer_deal_versioned_buffer(&(writer)->fw, &wb->bf, wb->version)
+
+static inline int flush_writer_files(SFOrderedWriterInfo *writer)
+{
+    int result;
+
+    if ((result=sf_file_writer_flush(&writer->fw)) != 0) {
+        return result;
+    }
+
+    if (writer->fw.flags & SF_FILE_WRITER_FLAGS_WANT_DONE_VERSION) {
+        writer->fw.last_versions.done = writer->fw.last_versions.pending;
+    }
+
+    return 0;
+}
+
+static inline int deal_versioned_binlog(SFOrderedWriterContext *context)
+{
+    SFOrderedWriterBuffer *wb;
+    int result;
+
+    while (1) {
+        if ((wb=sorted_queue_pop(&context->thread.queues.buffer,
+                        &context->thread.waiting)) != NULL)
+        {
+            context->writer.fw.total_count++;
+            result = deal_binlog_one_record(&context->writer, wb);
+            fast_mblock_free_object(&context->thread.allocators.buffer, wb);
+            return result;
+        }
+    }
+
+    return 0;
+}
+
+static int deal_version_chain(SFOrderedWriterContext *context,
+        struct fc_queue_info *qinfo)
+{
+    int result;
+    SFWriterVersionEntry *current_ver;
+    struct fast_mblock_node *prev_node;
+    struct fast_mblock_node *curr_node;
+    struct fast_mblock_chain node_chain;
+
+    current_ver = qinfo->head;
+    prev_node = NULL;
+    do {
+        curr_node = fast_mblock_to_node_ptr(current_ver);
+        if (prev_node != NULL) {
+            prev_node->next = curr_node;
+        }
+        prev_node = curr_node;
+
+        context->thread.waiting.version = current_ver->version;
+        if ((result=deal_versioned_binlog(context)) != 0) {
+            return result;
+        }
+    } while ((current_ver=current_ver->next) != NULL);
+
+    node_chain.head = fast_mblock_to_node_ptr(qinfo->head);
+    node_chain.tail = prev_node;
+    prev_node->next = NULL;
+    fast_mblock_batch_free(&context->thread.allocators.version, &node_chain);
+    return flush_writer_files(&context->writer);
+}
+
+void sf_ordered_writer_finish(SFOrderedWriterContext *ctx)
+{
+    int count;
+
+    if (ctx->writer.fw.file.name.str != NULL) {
+        fc_queue_terminate(&ctx->thread.queues.version);
+
+        count = 0;
+        while (ctx->thread.running && ++count < 300) {
+            fc_sleep_ms(10);
+        }
+
+        if (ctx->thread.running) {
+            logWarning("file: "__FILE__", line: %d, "
+                    "%s binlog write thread still running, exit anyway!",
+                    __LINE__, ctx->writer.fw.cfg.subdir_name);
+        }
+
+        free(ctx->writer.fw.file.name.str);
+        ctx->writer.fw.file.name.str = NULL;
+    }
+
+    if (ctx->writer.fw.file.fd >= 0) {
+        close(ctx->writer.fw.file.fd);
+        ctx->writer.fw.file.fd = -1;
+    }
+}
+
+static void *binlog_writer_func(void *arg)
+{
+    SFOrderedWriterContext *context;
+    SFOrderedWriterThread *thread;
+    struct fc_queue_info qinfo;
+
+    context = (SFOrderedWriterContext *)arg;
+    thread = &context->thread;
+
+#ifdef OS_LINUX
+    {
+        char thread_name[64];
+        fc_combine_two_strings(thread->name, "writer", '-', thread_name);
+        prctl(PR_SET_NAME, thread_name);
+    }
+#endif
+
+    thread->running = true;
+    while (SF_G_CONTINUE_FLAG) {
+        fc_queue_pop_to_queue(&thread->queues.version, &qinfo);
+        if (qinfo.head== NULL) {
+            continue;
+        }
+
+        if (deal_version_chain(context, &qinfo) != 0) {
+            logCrit("file: "__FILE__", line: %d, "
+                    "deal_version_chain fail, "
+                    "program exit!", __LINE__);
+            sf_terminate_myself();
+        }
+    }
+
+    thread->running = false;
+    return NULL;
+}
+
+static int binlog_wbuffer_alloc_init(void *element, void *args)
+{
+    SFOrderedWriterBuffer *wbuffer;
+    SFOrderedWriterInfo *writer;
+
+    wbuffer = (SFOrderedWriterBuffer *)element;
+    writer = (SFOrderedWriterInfo *)args;
+    wbuffer->bf.alloc_size = writer->fw.cfg.max_record_size;
+    wbuffer->bf.buff = (char *)(wbuffer + 1);
+    return 0;
+}
+
+static int push_compare_buffer_version(const SFOrderedWriterBuffer *entry1,
+        const SFOrderedWriterBuffer *entry2)
+{
+    return fc_compare_int64(entry1->version, entry2->version);
+}
+
+static int pop_compare_buffer_version(const SFOrderedWriterBuffer *entry,
+        const SFOrderedWriterBuffer *less_equal, void *arg)
+{
+    return fc_compare_int64(entry->version, less_equal->version);
+}
+
+static int sf_ordered_writer_init_thread(SFOrderedWriterContext *context,
+        const char *name, const int max_record_size)
+{
+    const int alloc_elements_once = 1024;
+    SFOrderedWriterThread *thread;
+    SFOrderedWriterInfo *writer;
+    int element_size;
+    pthread_t tid;
+    int result;
+
+    thread = &context->thread;
+    writer = &context->writer;
+    fc_safe_strcpy(thread->name, name);
+    writer->fw.cfg.max_record_size = max_record_size;
+    writer->thread = thread;
+
+    if ((result=fast_mblock_init_ex1(&thread->allocators.version,
+                    "writer-ver-info", sizeof(SFWriterVersionEntry),
+                    8 * 1024, 0, NULL, NULL, true)) != 0)
+    {
+        return result;
+    }
+
+    element_size = sizeof(SFOrderedWriterBuffer) + max_record_size;
+    if ((result=fast_mblock_init_ex1(&thread->allocators.buffer,
+                    "sorted-wbuffer", element_size, alloc_elements_once,
+                    0, binlog_wbuffer_alloc_init, writer, true)) != 0)
+    {
+        return result;
+    }
+
+    if ((result=fc_queue_init(&thread->queues.version, (unsigned long)
+                    (&((SFWriterVersionEntry *)NULL)->next))) != 0)
+    {
+        return result;
+    }
+
+    if ((result=sorted_queue_init(&thread->queues.buffer, (unsigned long)
+                    (&((SFOrderedWriterBuffer *)NULL)->dlink),
+                    (int (*)(const void *, const void *))
+                    push_compare_buffer_version,
+                    (int (*)(const void *, const void *, void *arg))
+                    pop_compare_buffer_version, NULL)) != 0)
+    {
+        return result;
+    }
+
+    return fc_create_thread(&tid, binlog_writer_func,
+            context, SF_G_THREAD_STACK_SIZE);
+}
+
+int sf_ordered_writer_init_ex(SFOrderedWriterContext *context,
+        const char *data_path, const char *subdir_name,
+        const char *file_prefix, const int buffer_size,
+        const int max_record_size, const int64_t file_rotate_size,
+        const bool call_fsync)
+{
+    int result;
+    if ((result=sf_file_writer_init(&context->writer.fw, data_path,
+                    subdir_name, file_prefix, max_record_size,
+                    buffer_size, file_rotate_size, call_fsync)) != 0)
+    {
+        return result;
+    }
+
+    return sf_ordered_writer_init_thread(context,
+            subdir_name, max_record_size);
+}

src/sf_ordered_writer.h

@@ -0,0 +1,142 @@
+/*
+ * Copyright (c) 2020 YuQing <384681@qq.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <https://www.gnu.org/licenses/>.
+ */
+
+//sf_ordered_writer.h
+
+#ifndef _SF_ORDERED_WRITER_H_
+#define _SF_ORDERED_WRITER_H_
+
+#include "fastcommon/sorted_queue.h"
+#include "sf_file_writer.h"
+
+typedef struct sf_writer_version_entry {
+    int64_t version;
+    struct sf_writer_version_entry *next;
+} SFWriterVersionEntry;
+
+typedef struct sf_ordered_writer_buffer {
+    int64_t version;
+    BufferInfo bf;
+    struct fc_list_head dlink;
+} SFOrderedWriterBuffer;
+
+typedef struct sf_orderd_writer_thread {
+    struct {
+        struct fast_mblock_man version;
+        struct fast_mblock_man buffer;
+    } allocators;
+
+    struct {
+        struct fc_queue version;
+        struct sorted_queue buffer;
+    } queues;
+    char name[64];
+    volatile bool running;
+    SFOrderedWriterBuffer waiting; //for less equal than object
+} SFOrderedWriterThread;
+
+typedef struct sf_ordered_writer_info {
+    SFFileWriterInfo fw;
+    SFBinlogBuffer binlog_buffer;
+    SFOrderedWriterThread *thread;
+} SFOrderedWriterInfo;
+
+typedef struct sf_ordered_writer_context {
+    SFOrderedWriterInfo writer;
+    SFOrderedWriterThread thread;
+} SFOrderedWriterContext;
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+int sf_ordered_writer_init_ex(SFOrderedWriterContext *context,
+        const char *data_path, const char *subdir_name,
+        const char *file_prefix, const int buffer_size,
+        const int max_record_size, const int64_t file_rotate_size,
+        const bool call_fsync);
+
+#define sf_ordered_writer_init(context, data_path, \
+        subdir_name, buffer_size, max_record_size) \
+    sf_ordered_writer_init_ex(context, data_path, subdir_name,   \
+            SF_BINLOG_FILE_PREFIX_STR, buffer_size, max_record_size, \
+            SF_BINLOG_DEFAULT_ROTATE_SIZE, true)
+
+#define sf_ordered_writer_set_flags(ctx, flags) \
+    sf_file_writer_set_flags(&(ctx)->writer.fw, flags)
+
+#define sf_ordered_writer_get_last_version(ctx) \
+    sf_ordered_writer_get_last_version(&(ctx)->writer.fw)
+
+void sf_ordered_writer_finish(SFOrderedWriterContext *ctx);
+
+#define sf_ordered_writer_get_current_index(ctx) \
+    sf_file_writer_get_current_index(&(ctx)->writer.fw)
+
+#define sf_ordered_writer_get_current_position(ctx, position) \
+    sf_file_writer_get_current_position(&(ctx)->writer.fw, position)
+
+static inline int sf_ordered_writer_alloc_versions(
+        SFOrderedWriterContext *ctx, const int count,
+        struct fc_queue_info *chain)
+{
+    return fc_queue_alloc_chain(&ctx->thread.queues.version,
+            &ctx->thread.allocators.version, count, chain);
+}
+
+static inline void sf_ordered_writer_push_versions(
+        SFOrderedWriterContext *ctx, struct fc_queue_info *chain)
+{
+    fc_queue_push_queue_to_tail(&ctx->thread.queues.version, chain);
+}
+
+static inline SFOrderedWriterBuffer *sf_ordered_writer_alloc_buffer(
+        SFOrderedWriterContext *ctx, const int64_t version)
+{
+    SFOrderedWriterBuffer *buffer;
+    buffer = (SFOrderedWriterBuffer *)fast_mblock_alloc_object(
+            &ctx->thread.allocators.buffer);
+    if (buffer != NULL) {
+        buffer->version = version;
+    }
+    return buffer;
+}
+
+#define sf_ordered_writer_get_filepath(data_path, subdir_name, filename, size) \
+    sf_file_writer_get_filepath(data_path, subdir_name, filename, size)
+
+#define sf_ordered_writer_get_filename(data_path,  \
+        subdir_name, binlog_index, filename, size) \
+        sf_file_writer_get_filename(data_path, subdir_name, \
+                binlog_index, filename, size)
+
+#define sf_ordered_writer_set_binlog_index(ctx, binlog_index) \
+    sf_file_writer_set_binlog_index(&(ctx)->writer.fw, binlog_index)
+
+#define sf_ordered_writer_push_to_thread_queue(ctx, buffer) \
+    sorted_queue_push(&(ctx)->thread.queues.buffer, buffer)
+
+static inline void sf_ordered_writer_push_to_queue(
+        SFOrderedWriterContext *ctx,
+        SFOrderedWriterBuffer *buffer)
+{
+    sorted_queue_push(&ctx->thread.queues.buffer, buffer);
+}
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif

src/sf_proto.c

@@ -14,27 +14,36 @@
  */
 
 
-#include <errno.h>
-#include "fastcommon/shared_func.h"
 #include "sf_util.h"
+#include "sf_nio.h"
 #include "sf_proto.h"
 
+static SFHandlerContext sf_handler_ctx = {NULL, {NULL, NULL}};
+static int64_t log_slower_than_us = 0;
+
+#define GET_CMD_CAPTION(cmd)    sf_handler_ctx.callbacks.get_cmd_caption(cmd)
+#define GET_CMD_LOG_LEVEL(cmd)  sf_handler_ctx.callbacks.get_cmd_log_level(cmd)
+
 int sf_proto_set_body_length(struct fast_task_info *task)
 {
     SFCommonProtoHeader *header;
+    char formatted_ip[FORMATTED_IP_SIZE];
 
-    header = (SFCommonProtoHeader *)task->data;
+    header = (SFCommonProtoHeader *)task->recv.ptr->data;
     if (!SF_PROTO_CHECK_MAGIC(header->magic)) {
+        format_ip_address(task->client_ip, formatted_ip);
         logError("file: "__FILE__", line: %d, "
-                "peer %s:%u, magic "SF_PROTO_MAGIC_FORMAT
-                " is invalid, expect: "SF_PROTO_MAGIC_FORMAT,
-                __LINE__, task->client_ip, task->port,
+                "%s peer %s:%u, magic "SF_PROTO_MAGIC_FORMAT" is invalid, "
+                "expect: "SF_PROTO_MAGIC_FORMAT", cmd: %d, body length: %d",
+                __LINE__, (task->handler != NULL ? task->handler->fh->ctx->
+                    name : ""), formatted_ip, task->port,
                 SF_PROTO_MAGIC_PARAMS(header->magic),
-                SF_PROTO_MAGIC_EXPECT_PARAMS);
+                SF_PROTO_MAGIC_EXPECT_PARAMS, header->cmd,
+                buff2int(header->body_len));
         return EINVAL;
     }
 
-    task->length = buff2int(header->body_len); //set body length
+    task->recv.ptr->length = buff2int(header->body_len); //set body length
     return 0;
 }
 
@@ -63,8 +72,14 @@ int sf_check_response(ConnectionInfo *conn, SFResponseInfo *response,
             response->error.length = response->header.body_len;
         }
 
-        if ((result=tcprecvdata_nb_ex(conn->sock, response->error.message,
-                response->error.length, network_timeout, &recv_bytes)) == 0)
+        if (conn->comm_type == fc_comm_type_rdma) {
+            memcpy(response->error.message, G_RDMA_CONNECTION_CALLBACKS.
+                    get_recv_buffer(conn)->buff + sizeof(SFCommonProtoHeader),
+                    response->error.length);
+            response->error.message[response->error.length] = '\0';
+        } else if ((result=tcprecvdata_nb_ex(conn->sock, response->
+                        error.message, response->error.length,
+                        network_timeout, &recv_bytes)) == 0)
         {
             response->error.message[response->error.length] = '\0';
         } else {
@@ -85,32 +100,66 @@ int sf_check_response(ConnectionInfo *conn, SFResponseInfo *response,
     return response->header.status;
 }
 
+static inline int sf_recv_response_header(ConnectionInfo *conn,
+        SFResponseInfo *response, const int network_timeout)
+{
+    int result;
+    BufferInfo *buffer;
+    SFCommonProtoHeader header_proto;
+
+    if (conn->comm_type == fc_comm_type_rdma) {
+        buffer = G_RDMA_CONNECTION_CALLBACKS.get_recv_buffer(conn);
+        if (buffer->length < sizeof(SFCommonProtoHeader)) {
+            response->error.length = sprintf(response->error.message,
+                    "recv pkg length: %d < header size: %d",
+                    buffer->length, (int)sizeof(SFCommonProtoHeader));
+            return EINVAL;
+        }
+
+        if ((result=sf_proto_parse_header((SFCommonProtoHeader *)
+                        buffer->buff, response)) != 0)
+        {
+            return result;
+        }
+
+        if (buffer->length != (sizeof(SFCommonProtoHeader) +
+                    response->header.body_len))
+        {
+            response->error.length = snprintf(response->error.message,
+                    sizeof(response->error.message),
+                    "recv package length: %d != calculate: %d",
+                    buffer->length, (int)(sizeof(SFCommonProtoHeader) +
+                        response->header.body_len));
+            return EINVAL;
+        }
+
+        return 0;
+    } else {
+        if ((result=tcprecvdata_nb(conn->sock, &header_proto,
+                        sizeof(SFCommonProtoHeader), network_timeout)) != 0)
+        {
+            response->error.length = snprintf(response->error.message,
+                    sizeof(response->error.message),
+                    "recv data fail, errno: %d, error info: %s",
+                    result, STRERROR(result));
+            return result;
+        }
+        return sf_proto_parse_header(&header_proto, response);
+    }
+}
+
 int sf_send_and_recv_response_header(ConnectionInfo *conn, char *data,
         const int len, SFResponseInfo *response, const int network_timeout)
 {
     int result;
-    SFCommonProtoHeader header_proto;
 
-    if ((result=tcpsenddata_nb(conn->sock, data, len, network_timeout)) != 0) {
-        response->error.length = snprintf(response->error.message,
-                sizeof(response->error.message),
-                "send data fail, errno: %d, error info: %s",
-                result, STRERROR(result));
-        return result;
-    }
-
-    if ((result=tcprecvdata_nb(conn->sock, &header_proto,
-            sizeof(SFCommonProtoHeader), network_timeout)) != 0)
+    if ((result=sf_proto_send_buf1(conn, data, len,
+                    response, network_timeout)) != 0)
     {
-        response->error.length = snprintf(response->error.message,
-                sizeof(response->error.message),
-                "recv data fail, errno: %d, error info: %s",
-                result, STRERROR(result));
         return result;
     }
 
-    sf_proto_extract_header(&header_proto, &response->header);
-    return 0;
+    return sf_recv_response_header(conn, response, network_timeout);
 }
 
 int sf_send_and_recv_response_ex(ConnectionInfo *conn, char *send_data,
@@ -169,7 +218,10 @@ int sf_send_and_recv_response_ex(ConnectionInfo *conn, char *send_data,
         return 0;
     }
 
-    if ((result=tcprecvdata_nb_ex(conn->sock, recv_data, response->
+    if (conn->comm_type == fc_comm_type_rdma) {
+        memcpy(recv_data, G_RDMA_CONNECTION_CALLBACKS.get_recv_buffer(conn)->
+                buff + sizeof(SFCommonProtoHeader), response->header.body_len);
+    } else if ((result=tcprecvdata_nb_ex(conn->sock, recv_data, response->
                     header.body_len, network_timeout, &recv_bytes)) != 0)
     {
         response->error.length = snprintf(response->error.message,
@@ -209,7 +261,11 @@ int sf_send_and_recv_response_ex1(ConnectionInfo *conn, char *send_data,
         return EOVERFLOW;
     }
 
-    if ((result=tcprecvdata_nb_ex(conn->sock, recv_data, response->
+    if (conn->comm_type == fc_comm_type_rdma) {
+        memcpy(recv_data, G_RDMA_CONNECTION_CALLBACKS.get_recv_buffer(conn)->
+                buff + sizeof(SFCommonProtoHeader), response->header.body_len);
+        *body_len = response->header.body_len;
+    } else if ((result=tcprecvdata_nb_ex(conn->sock, recv_data, response->
                     header.body_len, network_timeout, body_len)) != 0)
     {
         response->error.length = snprintf(response->error.message,
@@ -227,19 +283,12 @@ int sf_recv_response(ConnectionInfo *conn, SFResponseInfo *response,
 {
     int result;
     int recv_bytes;
-    SFCommonProtoHeader header_proto;
 
-    if ((result=tcprecvdata_nb(conn->sock, &header_proto,
-                    sizeof(SFCommonProtoHeader), network_timeout)) != 0)
+    if ((result=sf_recv_response_header(conn, response,
+                    network_timeout)) != 0)
     {
-        response->error.length = snprintf(response->error.message,
-                sizeof(response->error.message),
-                "recv data fail, errno: %d, error info: %s",
-                result, STRERROR(result));
         return result;
     }
-    sf_proto_extract_header(&header_proto, &response->header);
-
     if ((result=sf_check_response(conn, response, network_timeout,
                     expect_cmd)) != 0)
     {
@@ -257,8 +306,11 @@ int sf_recv_response(ConnectionInfo *conn, SFResponseInfo *response,
         return 0;
     }
 
-    if ((result=tcprecvdata_nb_ex(conn->sock, recv_data,
-                    expect_body_len, network_timeout, &recv_bytes)) != 0)
+    if (conn->comm_type == fc_comm_type_rdma) {
+        memcpy(recv_data, G_RDMA_CONNECTION_CALLBACKS.get_recv_buffer(conn)->
+                buff + sizeof(SFCommonProtoHeader), response->header.body_len);
+    } else if ((result=tcprecvdata_nb_ex(conn->sock, recv_data, expect_body_len,
+                    network_timeout, &recv_bytes)) != 0)
     {
         response->error.length = snprintf(response->error.message,
                 sizeof(response->error.message),
@@ -271,6 +323,94 @@ int sf_recv_response(ConnectionInfo *conn, SFResponseInfo *response,
     return result;
 }
 
+int sf_recv_vary_response(ConnectionInfo *conn, SFResponseInfo *response,
+        const int network_timeout, const unsigned char expect_cmd,
+        SFProtoRecvBuffer *buffer, const int min_body_len)
+{
+    int result;
+    int recv_bytes;
+
+    if ((result=sf_recv_response_header(conn, response,
+                    network_timeout)) != 0)
+    {
+        return result;
+    }
+    if ((result=sf_check_response(conn, response, network_timeout,
+                    expect_cmd)) != 0)
+    {
+        return result;
+    }
+
+    if (response->header.body_len < min_body_len) {
+        response->error.length = sprintf(response->error.message,
+                "response body length: %d < %d",
+                response->header.body_len, min_body_len);
+        return EINVAL;
+    }
+
+    if (response->header.body_len <= buffer->alloc_size) {
+        if (response->header.body_len == 0) {
+            return 0;
+        }
+    } else {
+        int alloc_size;
+        char *buff;
+
+        if (buffer->alloc_size > 0) {
+            alloc_size = 2 * buffer->alloc_size;
+        } else {
+            alloc_size = 64 * 1024;
+        }
+        while (alloc_size < response->header.body_len) {
+            alloc_size *= 2;
+        }
+
+        buff = (char *)fc_malloc(alloc_size);
+        if (buff == NULL) {
+            return ENOMEM;
+        }
+        if (buffer->buff != buffer->fixed && buffer->buff != NULL) {
+            free(buffer->buff);
+        }
+
+        buffer->buff = buff;
+        buffer->alloc_size = alloc_size;
+    }
+
+    if (conn->comm_type == fc_comm_type_rdma) {
+        memcpy(buffer->buff, G_RDMA_CONNECTION_CALLBACKS.get_recv_buffer(conn)->
+                buff + sizeof(SFCommonProtoHeader), response->header.body_len);
+    } else if ((result=tcprecvdata_nb_ex(conn->sock, buffer->buff, response->
+                    header.body_len, network_timeout, &recv_bytes)) != 0)
+    {
+        response->error.length = snprintf(response->error.message,
+                sizeof(response->error.message),
+                "recv body fail, recv bytes: %d, expect body length: %d, "
+                "errno: %d, error info: %s", recv_bytes,
+                response->header.body_len,
+                result, STRERROR(result));
+    }
+
+    return result;
+}
+
+int sf_send_and_recv_vary_response(ConnectionInfo *conn,
+        char *send_data, const int send_len, SFResponseInfo *response,
+        const int network_timeout, const unsigned char expect_cmd,
+        SFProtoRecvBuffer *buffer, const int min_body_len)
+{
+    int result;
+
+    if ((result=sf_proto_send_buf1(conn, send_data, send_len,
+                    response, network_timeout)) != 0)
+    {
+        return result;
+    }
+
+    return sf_recv_vary_response(conn, response, network_timeout,
+            expect_cmd, buffer, min_body_len);
+}
+
 const char *sf_get_cmd_caption(const int cmd)
 {
     switch (cmd) {
@@ -288,10 +428,26 @@ const char *sf_get_cmd_caption(const int cmd)
             return "CLOSE_CHANNEL_REQ";
         case SF_SERVICE_PROTO_CLOSE_CHANNEL_RESP:
             return "CLOSE_CHANNEL_RESP";
+        case SF_SERVICE_PROTO_REBIND_CHANNEL_REQ:
+            return "REBIND_CHANNEL_REQ";
+        case SF_SERVICE_PROTO_REBIND_CHANNEL_RESP:
+            return "REBIND_CHANNEL_RESP";
         case SF_SERVICE_PROTO_REPORT_REQ_RECEIPT_REQ:
             return "REPORT_REQ_RECEIPT_REQ";
         case SF_SERVICE_PROTO_REPORT_REQ_RECEIPT_RESP:
             return "REPORT_REQ_RECEIPT_RESP";
+        case SF_SERVICE_PROTO_GET_GROUP_SERVERS_REQ:
+            return "GET_GROUP_SERVERS_REQ";
+        case SF_SERVICE_PROTO_GET_GROUP_SERVERS_RESP:
+            return "GET_GROUP_SERVERS_RESP";
+        case SF_SERVICE_PROTO_GET_LEADER_REQ:
+            return "GET_LEADER_REQ";
+        case SF_SERVICE_PROTO_GET_LEADER_RESP:
+            return "GET_LEADER_RESP";
+        case SF_CLUSTER_PROTO_GET_SERVER_STATUS_REQ:
+            return "GET_SERVER_STATUS_REQ";
+        case SF_CLUSTER_PROTO_GET_SERVER_STATUS_RESP:
+            return "GET_SERVER_STATUS_RESP";
         default:
             return "UNKOWN";
     }
@@ -300,14 +456,17 @@ const char *sf_get_cmd_caption(const int cmd)
 int sf_proto_deal_ack(struct fast_task_info *task,
         SFRequestInfo *request, SFResponseInfo *response)
 {
+    char formatted_ip[FORMATTED_IP_SIZE];
+
     if (request->header.status != 0) {
         if (request->header.body_len > 0) {
             int remain_size;
             int len;
 
+            format_ip_address(task->client_ip, formatted_ip);
             response->error.length = sprintf(response->error.message,
                     "message from peer %s:%u => ",
-                    task->client_ip, task->port);
+                    formatted_ip, task->port);
             remain_size = sizeof(response->error.message) -
                 response->error.length;
             if (request->header.body_len >= remain_size) {
@@ -333,3 +492,238 @@ int sf_proto_deal_ack(struct fast_task_info *task,
 
     return 0;
 }
+
+int sf_proto_rebind_idempotency_channel(ConnectionInfo *conn,
+        const char *service_name, const uint32_t channel_id,
+        const int key, const int network_timeout)
+{
+    char out_buff[sizeof(SFCommonProtoHeader) +
+        sizeof(SFProtoRebindChannelReq)];
+    SFCommonProtoHeader *header;
+    SFProtoRebindChannelReq *req;
+    SFResponseInfo response;
+    int result;
+
+    header = (SFCommonProtoHeader *)out_buff;
+    req = (SFProtoRebindChannelReq *)(header + 1);
+    int2buff(channel_id, req->channel_id);
+    int2buff(key, req->key);
+    SF_PROTO_SET_HEADER(header, SF_SERVICE_PROTO_REBIND_CHANNEL_REQ,
+            sizeof(SFProtoRebindChannelReq));
+    response.error.length = 0;
+    if ((result=sf_send_and_recv_none_body_response(conn, out_buff,
+                    sizeof(out_buff), &response, network_timeout,
+                    SF_SERVICE_PROTO_REBIND_CHANNEL_RESP)) != 0)
+    {
+        sf_log_network_error(&response, conn, service_name, result);
+    }
+
+    return result;
+}
+
+int sf_proto_get_group_servers(ConnectionInfo *conn,
+        const char *service_name, const int network_timeout,
+        const int group_id, SFGroupServerArray *sarray)
+{
+    char out_buff[sizeof(SFCommonProtoHeader) +
+        sizeof(SFProtoGetGroupServersReq)];
+    char in_buff[1024];
+    char formatted_ip[FORMATTED_IP_SIZE];
+    SFCommonProtoHeader *header;
+    SFProtoGetGroupServersReq *req;
+    SFProtoGetGroupServersRespBodyHeader *body_header;
+    SFProtoGetGroupServersRespBodyPart *body_part;
+    SFGroupServerInfo *server;
+    SFGroupServerInfo *end;
+    SFResponseInfo response;
+    int result;
+    int body_len;
+    int count;
+
+    header = (SFCommonProtoHeader *)out_buff;
+    req = (SFProtoGetGroupServersReq *)(header + 1);
+    int2buff(group_id, req->group_id);
+    SF_PROTO_SET_HEADER(header, SF_SERVICE_PROTO_GET_GROUP_SERVERS_REQ,
+            sizeof(SFProtoGetGroupServersReq));
+    response.error.length = 0;
+    if ((result=sf_send_and_recv_response_ex1(conn, out_buff,
+                    sizeof(out_buff), &response, network_timeout,
+                    SF_SERVICE_PROTO_GET_GROUP_SERVERS_RESP, in_buff,
+                    sizeof(in_buff), &body_len)) != 0)
+    {
+        sf_log_network_error(&response, conn, service_name, result);
+        return result;
+    }
+
+    if (body_len < sizeof(SFProtoGetGroupServersRespBodyHeader)) {
+        format_ip_address(conn->ip_addr, formatted_ip);
+        logError("file: "__FILE__", line: %d, "
+                "server %s:%u response body length: %d < %d",
+                __LINE__, formatted_ip, conn->port, body_len,
+                (int)sizeof(SFProtoGetGroupServersRespBodyHeader));
+        return EINVAL;
+    }
+
+    body_header = (SFProtoGetGroupServersRespBodyHeader *)in_buff;
+    count = buff2int(body_header->count);
+    if (count <= 0) {
+        format_ip_address(conn->ip_addr, formatted_ip);
+        logError("file: "__FILE__", line: %d, "
+                "server %s:%u response server count: %d <= 0",
+                __LINE__, formatted_ip, conn->port, count);
+        return EINVAL;
+    }
+    if (count > sarray->alloc) {
+        format_ip_address(conn->ip_addr, formatted_ip);
+        logError("file: "__FILE__", line: %d, "
+                "server %s:%u response server count: %d is too large, "
+                "exceeds %d", __LINE__, formatted_ip, conn->port,
+                count, sarray->alloc);
+        return EOVERFLOW;
+    }
+    sarray->count = count;
+
+    body_part = (SFProtoGetGroupServersRespBodyPart *)(body_header + 1);
+    end = sarray->servers + sarray->count;
+    for (server=sarray->servers; server<end; server++, body_part++) {
+        server->id = buff2int(body_part->server_id);
+        server->is_master = body_part->is_master;
+        server->is_active = body_part->is_active;
+    }
+
+    return 0;
+}
+
+int sf_proto_get_leader(ConnectionInfo *conn, const char *service_name,
+        const int network_timeout, SFClientServerEntry *leader)
+{
+    int result;
+    SFCommonProtoHeader *header;
+    SFResponseInfo response;
+    SFProtoGetServerResp server_resp;
+    char out_buff[sizeof(SFCommonProtoHeader)];
+
+    header = (SFCommonProtoHeader *)out_buff;
+    SF_PROTO_SET_HEADER(header, SF_SERVICE_PROTO_GET_LEADER_REQ,
+            sizeof(out_buff) - sizeof(SFCommonProtoHeader));
+    if ((result=sf_send_and_recv_response(conn, out_buff,
+                    sizeof(out_buff), &response, network_timeout,
+                    SF_SERVICE_PROTO_GET_LEADER_RESP, (char *)&server_resp,
+                    sizeof(SFProtoGetServerResp))) != 0)
+    {
+        sf_log_network_error(&response, conn, service_name, result);
+    } else {
+        leader->server_id = buff2int(server_resp.server_id);
+        memcpy(leader->conn.ip_addr, server_resp.ip_addr, IP_ADDRESS_SIZE);
+        *(leader->conn.ip_addr + IP_ADDRESS_SIZE - 1) = '\0';
+        leader->conn.port = buff2short(server_resp.port);
+        leader->conn.comm_type = conn->comm_type;
+    }
+
+    return result;
+}
+
+void sf_proto_set_handler_context(const SFHandlerContext *ctx)
+{
+    sf_handler_ctx = *ctx;
+    log_slower_than_us = ctx->slow_log->cfg.log_slower_than_ms * 1000;
+}
+
+int sf_proto_deal_task_done(struct fast_task_info *task,
+        const char *service_name, SFCommonTaskContext *ctx)
+{
+    SFCommonProtoHeader *proto_header;
+    int status;
+    int r;
+    int64_t time_used;
+    int log_level;
+    char formatted_ip[FORMATTED_IP_SIZE];
+    char time_buff[32];
+
+    if (ctx->log_level != LOG_NOTHING && ctx->response.error.length > 0) {
+        log_it_ex(&g_log_context, ctx->log_level,
+                "file: "__FILE__", line: %d, %s "
+                "peer %s:%u, cmd: %d (%s), req body length: %d, "
+                "resp status: %d, %s", __LINE__, service_name,
+                format_ip_address(task->client_ip, formatted_ip),
+                task->port, ctx->request.header.cmd,
+                GET_CMD_CAPTION(ctx->request.header.cmd),
+                ctx->request.header.body_len, ctx->response.header.status,
+                ctx->response.error.message);
+    }
+
+    if (!ctx->need_response) {
+        if (sf_handler_ctx.callbacks.get_cmd_log_level != NULL) {
+            time_used = get_current_time_us() - ctx->req_start_time;
+            log_level = GET_CMD_LOG_LEVEL(ctx->request.header.cmd);
+            log_it_ex(&g_log_context, log_level, "file: "__FILE__", line: %d, "
+                    "%s client %s:%u, req cmd: %d (%s), req body_len: %d, "
+                    "resp status: %d, time used: %s us", __LINE__, service_name,
+                    format_ip_address(task->client_ip, formatted_ip),
+                    task->port, ctx->request.header.cmd,
+                    GET_CMD_CAPTION(ctx->request.header.cmd),
+                    ctx->request.header.body_len, ctx->response.header.status,
+                    long_to_comma_str(time_used, time_buff));
+        }
+
+        if (ctx->response.header.status == 0) {
+            return sf_set_read_event(task);
+        } else {
+            return FC_NEGATIVE(ctx->response.header.status);
+        }
+    }
+
+    proto_header = (SFCommonProtoHeader *)task->send.ptr->data;
+    if (!ctx->response_done) {
+        ctx->response.header.body_len = ctx->response.error.length;
+        if (ctx->response.error.length > 0) {
+            memcpy(task->send.ptr->data + sizeof(SFCommonProtoHeader),
+                    ctx->response.error.message, ctx->response.error.length);
+        }
+    }
+
+    status = sf_unify_errno(FC_ABS(ctx->response.header.status));
+    short2buff(status, proto_header->status);
+    short2buff(ctx->response.header.flags, proto_header->flags);
+    proto_header->cmd = ctx->response.header.cmd;
+    int2buff(ctx->response.header.body_len, proto_header->body_len);
+    task->send.ptr->length = sizeof(SFCommonProtoHeader) +
+        ctx->response.header.body_len;
+
+    r = sf_send_add_event(task);
+    time_used = get_current_time_us() - ctx->req_start_time;
+    if ((sf_handler_ctx.slow_log != NULL) && (sf_handler_ctx.slow_log->
+                cfg.enabled && time_used > log_slower_than_us))
+    {
+        char buff[256];
+        int blen;
+
+        blen = sprintf(buff, "timed used: %s us, %s client %s:%u, "
+                "req cmd: %d (%s), req body len: %d, resp cmd: %d (%s), "
+                "status: %d, resp body len: %d", long_to_comma_str(time_used,
+                    time_buff), service_name, format_ip_address(task->
+                        client_ip, formatted_ip), task->port, ctx->request.
+                header.cmd, GET_CMD_CAPTION(ctx->request.header.cmd),
+                ctx->request.header.body_len, ctx->response.header.cmd,
+                GET_CMD_CAPTION(ctx->response.header.cmd),
+                ctx->response.header.status, ctx->response.header.body_len);
+        log_it_ex2(&sf_handler_ctx.slow_log->ctx, NULL, buff, blen, false, true);
+    }
+
+    if (sf_handler_ctx.callbacks.get_cmd_log_level != NULL) {
+        log_level = GET_CMD_LOG_LEVEL(ctx->request.header.cmd);
+        log_it_ex(&g_log_context, log_level, "file: "__FILE__", line: %d, "
+                "%s client %s:%u, req cmd: %d (%s), req body_len: %d, "
+                "resp cmd: %d (%s), status: %d, resp body_len: %d, "
+                "time used: %s us", __LINE__, service_name,
+                format_ip_address(task->client_ip, formatted_ip),
+                task->port, ctx->request.header.cmd,
+                GET_CMD_CAPTION(ctx->request.header.cmd),
+                ctx->request.header.body_len, ctx->response.header.cmd,
+                GET_CMD_CAPTION(ctx->response.header.cmd),
+                ctx->response.header.status, ctx->response.header.body_len,
+                long_to_comma_str(time_used, time_buff));
+    }
+
+    return r == 0 ? ctx->response.header.status : r;
+}

src/sf_proto.h

@@ -25,6 +25,13 @@
 #include "fastcommon/sockopt.h"
 #include "sf_define.h"
 #include "sf_types.h"
+#include "sf_util.h"
+
+//for connection manager
+#define SF_SERVICE_PROTO_GET_GROUP_SERVERS_REQ    111
+#define SF_SERVICE_PROTO_GET_GROUP_SERVERS_RESP   112
+#define SF_SERVICE_PROTO_GET_LEADER_REQ           113
+#define SF_SERVICE_PROTO_GET_LEADER_RESP          114
 
 #define SF_PROTO_ACK                    116
 
@@ -32,13 +39,19 @@
 #define SF_PROTO_ACTIVE_TEST_RESP       118
 
 //for request idempotency
-#define SF_SERVICE_PROTO_SETUP_CHANNEL_REQ        121
-#define SF_SERVICE_PROTO_SETUP_CHANNEL_RESP       122
-#define SF_SERVICE_PROTO_CLOSE_CHANNEL_REQ        123
-#define SF_SERVICE_PROTO_CLOSE_CHANNEL_RESP       124
+#define SF_SERVICE_PROTO_SETUP_CHANNEL_REQ        119
+#define SF_SERVICE_PROTO_SETUP_CHANNEL_RESP       120
+#define SF_SERVICE_PROTO_CLOSE_CHANNEL_REQ        121
+#define SF_SERVICE_PROTO_CLOSE_CHANNEL_RESP       122
+#define SF_SERVICE_PROTO_REBIND_CHANNEL_REQ       123
+#define SF_SERVICE_PROTO_REBIND_CHANNEL_RESP      124
 #define SF_SERVICE_PROTO_REPORT_REQ_RECEIPT_REQ   125
 #define SF_SERVICE_PROTO_REPORT_REQ_RECEIPT_RESP  126
 
+#define SF_CLUSTER_PROTO_GET_SERVER_STATUS_REQ    201
+#define SF_CLUSTER_PROTO_GET_SERVER_STATUS_RESP   202
+
+
 #define SF_PROTO_MAGIC_CHAR        '@'
 #define SF_PROTO_SET_MAGIC(m)   \
     m[0] = m[1] = m[2] = m[3] = SF_PROTO_MAGIC_CHAR
@@ -55,19 +68,17 @@
 #define SF_PROTO_MAGIC_PARAMS(m) \
     m[0], m[1], m[2], m[3]
 
-#define SF_PROTO_SET_HEADER(header, _cmd, _body_len) \
+#define SF_PROTO_SET_HEADER_EX(header, _cmd, _flags, _body_len) \
     do {  \
         SF_PROTO_SET_MAGIC((header)->magic);   \
         (header)->cmd = _cmd;      \
         (header)->status[0] = (header)->status[1] = 0; \
+        short2buff(_flags, (header)->flags); \
         int2buff(_body_len, (header)->body_len); \
     } while (0)
 
-#define SF_PROTO_SET_HEADER_EX(header, _cmd, _flags, _body_len) \
-    do {  \
-        SF_PROTO_SET_HEADER(header, _cmd, _body_len); \
-        short2buff(_flags, (header)->flags); \
-    } while (0)
+#define SF_PROTO_SET_HEADER(header, _cmd, _body_len) \
+        SF_PROTO_SET_HEADER_EX(header, _cmd, 0, _body_len)
 
 #define SF_PROTO_SET_RESPONSE_HEADER(proto_header, resp_header) \
     do {  \
@@ -77,6 +88,52 @@
     } while (0)
 
 
+#define SF_PROTO_SEND_BODY(task)  \
+    (task->send.ptr->data + sizeof(SFCommonProtoHeader))
+
+#define SF_PROTO_RECV_BODY(task) \
+    (task->recv.ptr->data + sizeof(SFCommonProtoHeader))
+
+#define SF_RECV_BODY_LENGTH(task) \
+    (task->recv.ptr->length - sizeof(SFCommonProtoHeader))
+
+#define SF_SEND_BUFF_END(task) (task->send.ptr->data + task->send.ptr->size)
+#define SF_RECV_BUFF_END(task) (task->recv.ptr->data + task->recv.ptr->size)
+
+#define SF_PROTO_UPDATE_EXTRA_BODY_SIZE \
+    sizeof(SFProtoIdempotencyAdditionalHeader) + FCFS_AUTH_SESSION_ID_LEN
+
+#define SF_PROTO_QUERY_EXTRA_BODY_SIZE  FCFS_AUTH_SESSION_ID_LEN
+
+#define SF_PROTO_CLIENT_SET_REQ_EX(client_ctx, auth_enabled, \
+        out_buff, header, req, the_req_id, out_bytes) \
+    do {   \
+        char *the_req_start;  \
+        header = (SFCommonProtoHeader *)out_buff; \
+        the_req_start = (char *)(header + 1);     \
+        out_bytes = sizeof(SFCommonProtoHeader) + sizeof(*req); \
+        if (auth_enabled) { \
+            out_bytes += FCFS_AUTH_SESSION_ID_LEN;   \
+            memcpy(the_req_start, client_ctx->auth.ctx-> \
+                    session.id, FCFS_AUTH_SESSION_ID_LEN);  \
+            the_req_start += FCFS_AUTH_SESSION_ID_LEN;   \
+        }  \
+        if (the_req_id > 0) {  \
+            long2buff(the_req_id, ((SFProtoIdempotencyAdditionalHeader *)\
+                        the_req_start)->req_id);  \
+            out_bytes += sizeof(SFProtoIdempotencyAdditionalHeader); \
+            req = (typeof(req))(the_req_start +   \
+                    sizeof(SFProtoIdempotencyAdditionalHeader));     \
+        } else {  \
+            req = (typeof(req))the_req_start;  \
+        }  \
+    } while (0)
+
+#define SF_PROTO_CLIENT_SET_REQ(client_ctx, out_buff, \
+        header, req, the_req_id, out_bytes) \
+   SF_PROTO_CLIENT_SET_REQ_EX(client_ctx, client_ctx->auth.enabled, \
+           out_buff, header, req, the_req_id, out_bytes)
+
 typedef struct sf_common_proto_header {
     unsigned char magic[4]; //magic number
     char body_len[4];       //body length
@@ -86,6 +143,39 @@ typedef struct sf_common_proto_header {
     char padding[3];
 } SFCommonProtoHeader;
 
+typedef struct sf_proto_limit_info {
+    char offset[4];
+    char count[4];
+} SFProtoLimitInfo;
+
+typedef struct sf_proto_get_group_servers_req {
+    char group_id[4];
+    char padding[4];
+} SFProtoGetGroupServersReq;
+
+typedef struct sf_proto_get_group_servers_resp_body_header {
+    char count[4];
+    char padding[4];
+} SFProtoGetGroupServersRespBodyHeader;
+
+typedef struct sf_proto_get_group_servers_resp_body_part {
+    char server_id[4];
+    char is_master;
+    char is_active;
+    char padding[2];
+} SFProtoGetGroupServersRespBodyPart;
+
+typedef struct sf_proto_get_server_resp {
+    char ip_addr[IP_ADDRESS_SIZE];
+    char server_id[4];
+    char port[2];
+    char padding[2];
+} SFProtoGetServerResp;
+
+typedef struct sf_proto_empty_body_req {
+    char nothing[0];
+} SFProtoEmptyBodyReq;
+
 typedef struct sf_proto_idempotency_additional_header {
     char req_id[8];
 } SFProtoIdempotencyAdditionalHeader;
@@ -98,10 +188,15 @@ typedef struct sf_proto_setup_channel_req {
 typedef struct sf_proto_setup_channel_resp {
     char channel_id[4];
     char key[4];
+    char server_id[4];
     char buffer_size[4];
-    char padding[4];
 } SFProtoSetupChannelResp;
 
+typedef struct sf_proto_rebind_channel_req {
+    char channel_id[4];
+    char key[4];
+} SFProtoRebindChannelReq;
+
 typedef struct sf_proto_report_req_receipt_header {
     char count[4];
     char padding[4];
@@ -111,42 +206,281 @@ typedef struct sf_proto_report_req_receipt_body {
     char req_id[8];
 } SFProtoReportReqReceiptBody;
 
+typedef struct {
+    unsigned char servers[SF_CLUSTER_CONFIG_SIGN_LEN];
+    unsigned char cluster[SF_CLUSTER_CONFIG_SIGN_LEN];
+} SFProtoConfigSigns;
+
+typedef struct sf_proto_get_server_status_req {
+    SFProtoConfigSigns config_signs;
+    char server_id[4];  //my server id
+    union {
+        char is_leader;
+        char is_master;
+    };
+    char auth_enabled;
+    char padding[2];
+} SFProtoGetServerStatusReq;
+
+typedef struct sf_get_server_status_request {
+    const unsigned char *servers_sign;
+    const unsigned char *cluster_sign;
+    int server_id;  //my server id
+    union {
+        bool is_leader;
+        bool is_master;
+    };
+    bool auth_enabled;
+} SFGetServerStatusRequest;
+
+typedef struct sf_group_server_info {
+    int id;
+    bool is_leader;
+    bool is_master;
+    bool is_active;
+    char padding[1];
+} SFGroupServerInfo;
+
+typedef struct sf_group_server_array {
+    SFGroupServerInfo *servers;
+    int alloc;
+    int count;
+} SFGroupServerArray;
+
+typedef struct sf_client_server_entry {
+    int server_id;
+    ConnectionInfo conn;
+} SFClientServerEntry;
+
+typedef const char *(*sf_get_cmd_caption_func)(const int cmd);
+typedef int (*sf_get_cmd_log_level_func)(const int cmd);
+
+typedef struct {
+    int alloc_size;
+    int fixed_size;
+    char *fixed;
+    char *buff;
+} SFProtoRecvBuffer;
+
+typedef struct {
+    char fixed[64 * 1024];
+    SFProtoRecvBuffer buffer;
+} SFProtoRBufferFixedWrapper;
+
+typedef struct {
+    sf_get_cmd_caption_func get_cmd_caption;
+    sf_get_cmd_log_level_func get_cmd_log_level;
+} SFCommandCallbacks;
+
+typedef struct {
+    SFSlowLogContext *slow_log;
+    SFCommandCallbacks callbacks;
+} SFHandlerContext;
+
 #ifdef __cplusplus
 extern "C" {
 #endif
 
+void sf_proto_set_handler_context(const SFHandlerContext *ctx);
+
 int sf_proto_set_body_length(struct fast_task_info *task);
 
 const char *sf_get_cmd_caption(const int cmd);
 
+int sf_proto_deal_task_done(struct fast_task_info *task,
+        const char *service_name, SFCommonTaskContext *ctx);
+
+static inline void sf_proto_init_task_magic(struct fast_task_info *task)
+{
+    SF_PROTO_SET_MAGIC(((SFCommonProtoHeader *)
+                task->send.ptr->data)->magic);
+    if (task->recv.ptr != task->send.ptr) {
+        SF_PROTO_SET_MAGIC(((SFCommonProtoHeader *)
+                    task->recv.ptr->data)->magic);
+    }
+}
+
+static inline void sf_proto_init_task_context(struct fast_task_info *task,
+        SFCommonTaskContext *ctx)
+{
+    ctx->req_start_time = get_current_time_us();
+    ctx->response.header.cmd = SF_PROTO_ACK;
+    ctx->response.header.body_len = 0;
+    ctx->response.header.status = 0;
+    ctx->response.header.flags = 0;
+    ctx->response.error.length = 0;
+    ctx->response.error.message[0] = '\0';
+    ctx->log_level = LOG_ERR;
+    ctx->response_done = false;
+    ctx->need_response = true;
+
+    ctx->request.header.cmd = ((SFCommonProtoHeader *)
+            task->recv.ptr->data)->cmd;
+    ctx->request.header.body_len = SF_RECV_BODY_LENGTH(task);
+    ctx->request.header.status = buff2short(((SFCommonProtoHeader *)
+                task->recv.ptr->data)->status);
+    ctx->request.header.flags = buff2short(((SFCommonProtoHeader *)
+                task->recv.ptr->data)->flags);
+    if (task->recv_body != NULL) {
+        ctx->request.body = task->recv_body;
+    } else {
+        ctx->request.body = SF_PROTO_RECV_BODY(task);
+    }
+}
+
+/* task send and recv buffer operations */
+static inline int sf_set_task_send_buffer_size(
+        struct fast_task_info *task, const int expect_size)
+{
+    int result;
+    if ((result=free_queue_set_buffer_size(task, task->send.ptr,
+                    expect_size)) == 0)
+    {
+        SF_PROTO_SET_MAGIC(((SFCommonProtoHeader *)
+                    task->send.ptr->data)->magic);
+    }
+    return result;
+}
+
+static inline int sf_set_task_recv_buffer_size(
+        struct fast_task_info *task, const int expect_size)
+{
+    int result;
+    if ((result=free_queue_set_buffer_size(task, task->recv.ptr,
+                    expect_size)) == 0)
+    {
+        SF_PROTO_SET_MAGIC(((SFCommonProtoHeader *)
+                    task->recv.ptr->data)->magic);
+    }
+    return result;
+}
+
+static inline int sf_set_task_send_max_buffer_size(
+        struct fast_task_info *task)
+{
+    int result;
+    if ((result=free_queue_set_max_buffer_size(task, task->send.ptr)) == 0) {
+        SF_PROTO_SET_MAGIC(((SFCommonProtoHeader *)
+                    task->send.ptr->data)->magic);
+    }
+    return result;
+}
+
+static inline int sf_set_task_recv_max_buffer_size(
+        struct fast_task_info *task)
+{
+    int result;
+    if ((result=free_queue_set_max_buffer_size(task, task->recv.ptr)) == 0) {
+        SF_PROTO_SET_MAGIC(((SFCommonProtoHeader *)
+                    task->recv.ptr->data)->magic);
+    }
+    return result;
+}
+
+static inline int sf_realloc_task_send_buffer(
+        struct fast_task_info *task, const int expect_size)
+{
+    int result;
+    if ((result=free_queue_realloc_buffer(task, task->send.ptr,
+                    expect_size)) == 0)
+    {
+        SF_PROTO_SET_MAGIC(((SFCommonProtoHeader *)
+                    task->send.ptr->data)->magic);
+    }
+    return result;
+}
+
+static inline int sf_realloc_task_recv_buffer(
+        struct fast_task_info *task, const int expect_size)
+{
+    int result;
+    if ((result=free_queue_realloc_buffer(task, task->recv.ptr,
+                    expect_size)) == 0)
+    {
+        SF_PROTO_SET_MAGIC(((SFCommonProtoHeader *)
+                    task->recv.ptr->data)->magic);
+    }
+    return result;
+}
+
+static inline int sf_realloc_task_send_max_buffer(
+        struct fast_task_info *task)
+{
+    int result;
+    if ((result=free_queue_realloc_max_buffer(task, task->send.ptr)) == 0) {
+        SF_PROTO_SET_MAGIC(((SFCommonProtoHeader *)
+                    task->send.ptr->data)->magic);
+    }
+    return result;
+}
+
+static inline int sf_realloc_task_recv_max_buffer(
+        struct fast_task_info *task)
+{
+    int result;
+    if ((result=free_queue_realloc_max_buffer(task, task->recv.ptr)) == 0) {
+        SF_PROTO_SET_MAGIC(((SFCommonProtoHeader *)
+                    task->recv.ptr->data)->magic);
+    }
+    return result;
+}
+
+
 static inline void sf_log_network_error_ex1(SFResponseInfo *response,
-        const ConnectionInfo *conn, const int result,
-        const int log_level, const char *file, const int line)
+        const ConnectionInfo *conn, const char *service_name,
+        const int result, const int log_level,
+        const char *file, const int line)
 {
     if (response->error.length > 0) {
-        log_it_ex(&g_log_context, log_level,
-                "file: %s, line: %d, "
-                "server %s:%u, %s", file, line,
+        log_it_ex(&g_log_context, log_level, "file: %s, line: %d, "
+                "%s%sserver %s:%u response message: %s", file, line,
+                (service_name != NULL ? service_name : ""),
+                (service_name != NULL ? " ": ""),
                 conn->ip_addr, conn->port,
                 response->error.message);
     } else {
-        log_it_ex(&g_log_context, log_level,
-                "file: %s, line: %d, "
-                "communicate with server %s:%u fail, "
+        log_it_ex(&g_log_context, log_level, "file: %s, line: %d, "
+                "communicate with %s%sserver %s:%u fail, "
                 "errno: %d, error info: %s", file, line,
+                (service_name != NULL ? service_name : ""),
+                (service_name != NULL ? " ": ""),
                 conn->ip_addr, conn->port,
                 result, STRERROR(result));
     }
 }
 
-#define sf_log_network_error_ex(response, conn, result, log_level) \
-    sf_log_network_error_ex1(response, conn, result, \
-            log_level, __FILE__, __LINE__)
+#define sf_log_network_error_ex(response, conn, \
+        service_name, result, log_level) \
+    sf_log_network_error_ex1(response, conn, service_name, \
+            result, log_level, __FILE__, __LINE__)
 
-#define sf_log_network_error(response, conn, result) \
-    sf_log_network_error_ex1(response, conn, result, \
+#define sf_log_network_error(response, conn, service_name, result) \
+    sf_log_network_error_ex1(response, conn, service_name, result, \
             LOG_ERR, __FILE__, __LINE__)
 
+#define sf_log_network_error_for_update_ex(response, conn,  \
+        service_name, result, enoent_log_level, file, line) \
+        sf_log_network_error_ex1(response, conn, service_name, result, \
+                (result == SF_RETRIABLE_ERROR_CHANNEL_INVALID) ?       \
+                LOG_DEBUG : ((result == ENOENT || result == ENODATA) ? \
+                    enoent_log_level : LOG_ERR), file, line)
+
+#define sf_log_network_error_for_update(response, conn, service_name, result) \
+    sf_log_network_error_for_update_ex(response, conn, service_name, \
+            result, LOG_ERR, __FILE__, __LINE__)
+
+#define sf_log_network_error_for_delete_ex(response, conn,   \
+        service_name, result, enoent_log_level, file, line)  \
+        sf_log_network_error_ex1(response, conn, service_name, result,  \
+                (result == SF_RETRIABLE_ERROR_CHANNEL_INVALID) ? \
+                LOG_DEBUG : ((result == ENOENT || result == ENODATA) ? \
+                    enoent_log_level : LOG_ERR), file, line)
+
+#define sf_log_network_error_for_delete(response,      \
+        conn, service_name, result, enoent_log_level)  \
+        sf_log_network_error_for_delete_ex(response, conn, service_name, \
+                result, enoent_log_level, __FILE__, __LINE__)
+
 
 static inline int sf_server_expect_body_length(SFResponseInfo *response,
         const int body_length, const int expect_body_len)
@@ -204,6 +538,23 @@ static inline int sf_server_check_body_length(
             body_length, max_body_length);
 }
 
+#define server_expect_body_length(expect_body_len) \
+    sf_server_expect_body_length(&RESPONSE, REQUEST.header.body_len, \
+            expect_body_len)
+
+#define server_check_min_body_length(min_body_length) \
+    sf_server_check_min_body_length(&RESPONSE, REQUEST.header.body_len, \
+            min_body_length)
+
+#define server_check_max_body_length(max_body_length) \
+    sf_server_check_max_body_length(&RESPONSE, REQUEST.header.body_len, \
+            max_body_length)
+
+#define server_check_body_length(min_body_length, max_body_length) \
+    sf_server_check_body_length(&RESPONSE, REQUEST.header.body_len, \
+            min_body_length, max_body_length)
+
+
 int sf_check_response(ConnectionInfo *conn, SFResponseInfo *response,
         const int network_timeout, const unsigned char expect_cmd);
 
@@ -211,6 +562,100 @@ int sf_recv_response(ConnectionInfo *conn, SFResponseInfo *response,
         const int network_timeout, const unsigned char expect_cmd,
         char *recv_data, const int expect_body_len);
 
+static inline int sf_recv_none_body_response(ConnectionInfo *conn,
+        SFResponseInfo *response, const int network_timeout,
+        const unsigned char expect_cmd)
+{
+    char *recv_data = NULL;
+    const int expect_body_len = 0;
+    return sf_recv_response(conn, response, network_timeout,
+            expect_cmd, recv_data, expect_body_len);
+}
+
+int sf_recv_vary_response(ConnectionInfo *conn, SFResponseInfo *response,
+        const int network_timeout, const unsigned char expect_cmd,
+        SFProtoRecvBuffer *buffer, const int min_body_len);
+
+static inline void sf_init_recv_buffer_by_wrapper(
+        SFProtoRBufferFixedWrapper *wrapper)
+{
+    wrapper->buffer.fixed_size = sizeof(wrapper->fixed);
+    wrapper->buffer.alloc_size = sizeof(wrapper->fixed);
+    wrapper->buffer.fixed = wrapper->fixed;
+    wrapper->buffer.buff = wrapper->fixed;
+}
+
+static inline int sf_init_recv_buffer(SFProtoRecvBuffer *buffer,
+        const int init_size)
+{
+    buffer->alloc_size = init_size;
+    buffer->fixed_size = 0;
+    buffer->fixed = NULL;
+    buffer->buff = (char *)fc_malloc(init_size);
+    return buffer->buff != NULL ? 0 : ENOMEM;
+}
+
+static inline void sf_free_recv_buffer(SFProtoRecvBuffer *buffer)
+{
+    if (buffer->buff != buffer->fixed) {
+        if (buffer->buff != NULL) {
+            free(buffer->buff);
+        }
+        buffer->alloc_size = buffer->fixed_size;
+        buffer->buff = buffer->fixed;
+    }
+}
+
+static inline int sf_proto_send_buf1(ConnectionInfo *conn, char *data,
+        const int len, SFResponseInfo *response, const int network_timeout)
+{
+    int result;
+
+    if (conn->comm_type == fc_comm_type_rdma) {
+        result = G_RDMA_CONNECTION_CALLBACKS.request_by_buf1(
+                conn, data, len, network_timeout * 1000);
+    } else {
+        result = tcpsenddata_nb(conn->sock, data, len, network_timeout);
+    }
+    if (result != 0) {
+        response->error.length = snprintf(response->error.message,
+                sizeof(response->error.message),
+                "send data fail, errno: %d, error info: %s",
+                result, STRERROR(result));
+    }
+
+    return result;
+}
+
+static inline int sf_proto_send_buf2(ConnectionInfo *conn, char *buff1,
+        const int length1, char *buff2, const int length2,
+        SFResponseInfo *response, const int network_timeout)
+{
+    int result;
+
+    if (conn->comm_type == fc_comm_type_rdma) {
+        result = G_RDMA_CONNECTION_CALLBACKS.request_by_buf2(
+                conn, buff1, length1, buff2, length2,
+                network_timeout * 1000);
+    } else {
+        if ((result=tcpsenddata_nb(conn->sock, buff1, length1,
+                        network_timeout)) == 0)
+        {
+            result = tcpsenddata_nb(conn->sock, buff2, length2,
+                    network_timeout);
+        }
+    }
+
+    if (result != 0) {
+        response->error.length = snprintf(response->error.message,
+                sizeof(response->error.message),
+                "send data fail, errno: %d, error info: %s",
+                result, STRERROR(result));
+    }
+
+    return result;
+}
+
 int sf_send_and_recv_response_header(ConnectionInfo *conn, char *data,
         const int len, SFResponseInfo *response, const int network_timeout);
 
@@ -226,7 +671,6 @@ static inline int sf_send_and_check_response_header(ConnectionInfo *conn,
         return result;
     }
 
-
     if ((result=sf_check_response(conn, response, network_timeout,
                     expect_cmd)) != 0)
     {
@@ -267,13 +711,46 @@ static inline int sf_send_and_recv_none_body_response(ConnectionInfo *conn,
         network_timeout, expect_cmd, recv_data, expect_body_len);
 }
 
-static inline void sf_proto_extract_header(SFCommonProtoHeader *header_proto,
-        SFHeaderInfo *header_info)
+int sf_send_and_recv_vary_response(ConnectionInfo *conn,
+        char *send_data, const int send_len, SFResponseInfo *response,
+        const int network_timeout, const unsigned char expect_cmd,
+        SFProtoRecvBuffer *buffer, const int min_body_len);
+
+static inline int sf_proto_parse_header(const SFCommonProtoHeader
+        *header_proto, SFResponseInfo *response)
 {
-    header_info->cmd = header_proto->cmd;
-    header_info->body_len = buff2int(header_proto->body_len);
-    header_info->flags = buff2short(header_proto->flags);
-    header_info->status = buff2short(header_proto->status);
+    if (!SF_PROTO_CHECK_MAGIC(header_proto->magic)) {
+        response->error.length = snprintf(response->error.message,
+                sizeof(response->error.message),
+                "magic "SF_PROTO_MAGIC_FORMAT" is invalid, "
+                "expect: "SF_PROTO_MAGIC_FORMAT,
+                SF_PROTO_MAGIC_PARAMS(header_proto->magic),
+                SF_PROTO_MAGIC_EXPECT_PARAMS);
+        return EINVAL;
+    }
+
+    response->header.cmd = header_proto->cmd;
+    response->header.body_len = buff2int(header_proto->body_len);
+    response->header.flags = buff2short(header_proto->flags);
+    response->header.status = buff2short(header_proto->status);
+    if (response->header.status > 255) {
+        response->header.status = sf_localize_errno(response->header.status);
+    }
+    return 0;
+}
+
+static inline void sf_proto_pack_limit(const SFListLimitInfo
+        *limit_info, SFProtoLimitInfo *limit_proto)
+{
+    int2buff(limit_info->offset, limit_proto->offset);
+    int2buff(limit_info->count, limit_proto->count);
+}
+
+static inline void sf_proto_extract_limit(const SFProtoLimitInfo
+        *limit_proto, SFListLimitInfo *limit_info)
+{
+    limit_info->offset = buff2int(limit_proto->offset);
+    limit_info->count = buff2int(limit_proto->count);
 }
 
 static inline int sf_active_test(ConnectionInfo *conn,
@@ -297,16 +774,44 @@ static inline int sf_proto_deal_active_test(struct fast_task_info *task,
 int sf_proto_deal_ack(struct fast_task_info *task,
         SFRequestInfo *request, SFResponseInfo *response);
 
-#define SF_CLIENT_RELEASE_CONNECTION(client_ctx, conn, result) \
-    do {  \
+int sf_proto_rebind_idempotency_channel(ConnectionInfo *conn,
+        const char *service_name, const uint32_t channel_id,
+        const int key, const int network_timeout);
+
+int sf_proto_get_group_servers(ConnectionInfo *conn,
+        const char *service_name, const int network_timeout,
+        const int group_id, SFGroupServerArray *sarray);
+
+int sf_proto_get_leader(ConnectionInfo *conn, const char *service_name,
+        const int network_timeout, SFClientServerEntry *leader);
+
+static inline void sf_proto_get_server_status_pack(
+        const SFGetServerStatusRequest *r,
+        SFProtoGetServerStatusReq *req)
+{
+    int2buff(r->server_id, req->server_id);
+    req->is_leader = (r->is_leader ? 1 : 0);
+    req->auth_enabled = (r->auth_enabled ? 1 : 0);
+    memcpy(req->config_signs.servers, r->servers_sign,
+            SF_CLUSTER_CONFIG_SIGN_LEN);
+    if (r->cluster_sign != NULL) {
+        memcpy(req->config_signs.cluster, r->cluster_sign,
+                SF_CLUSTER_CONFIG_SIGN_LEN);
+    } else {
+        memset(req->config_signs.cluster, 0,
+                SF_CLUSTER_CONFIG_SIGN_LEN);
+    }
+}
+
+#define SF_CLIENT_RELEASE_CONNECTION(cm, conn, result) \
+    do { \
         if (SF_FORCE_CLOSE_CONNECTION_ERROR(result)) {  \
-            client_ctx->conn_manager.close_connection(client_ctx, conn);   \
-        } else if (client_ctx->conn_manager.release_connection != NULL) {  \
-            client_ctx->conn_manager.release_connection(client_ctx, conn); \
+            (cm)->ops.close_connection(cm, conn);   \
+        } else if ((cm)->ops.release_connection != NULL) {  \
+            (cm)->ops.release_connection(cm, conn); \
         } \
     } while (0)
 
-
 #ifdef __cplusplus
 }
 #endif

src/sf_serializer.c

@@ -0,0 +1,521 @@
+/*
+ * Copyright (c) 2020 YuQing <384681@qq.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <https://www.gnu.org/licenses/>.
+ */
+
+#include <time.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <ctype.h>
+#include <unistd.h>
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <netinet/in.h>
+#include <errno.h>
+#include "fastcommon/shared_func.h"
+#include "fastcommon/logger.h"
+#include "sf_serializer.h"
+
+#define FIELD_ID_AND_TYPE_FORMAT  "fid: %d, type: %s"
+#define FIELD_ID_AND_TYPE_PARAMS  it->field.fid, \
+    value_type_configs[it->field.type].name
+
+
+typedef struct {
+    const char *name;
+    int min_size;
+    int elt_size;
+} SFSerializerTypeConfig;
+
+static SFSerializerTypeConfig value_type_configs[SF_SERIALIZER_VALUE_TYPE_COUNT] =
+{
+    {"int8",   sizeof(SFSerializerPackFieldInt8),   0},
+    {"int16",  sizeof(SFSerializerPackFieldInt16),  0},
+    {"int32",  sizeof(SFSerializerPackFieldInt32),  0},
+    {"int64",  sizeof(SFSerializerPackFieldInt64),  0},
+    {"string", sizeof(SFSerializerPackStringValue), 0},
+    {"int8_array",  sizeof(SFSerializerPackFieldArray), 1},
+    {"int16_array", sizeof(SFSerializerPackFieldArray), 2},
+    {"int32_array", sizeof(SFSerializerPackFieldArray), 4},
+    {"int64_array", sizeof(SFSerializerPackFieldArray), 8},
+    {"string_array", sizeof(SFSerializerPackFieldArray),
+        sizeof(SFSerializerPackStringValue)},
+    {"id_name_array", sizeof(SFSerializerPackFieldArray),
+        sizeof(int64_t) + sizeof(SFSerializerPackStringValue)},
+    {"map", sizeof(SFSerializerPackFieldArray), 2 *
+        sizeof(SFSerializerPackStringValue)}
+};
+
+int sf_serializer_unpack(SFSerializerIterator *it, const string_t *content)
+{
+    SFSerializerPackHeader *header;
+    int length;
+    int calc_crc32;
+    int header_crc32;
+
+    if (content->len < sizeof(SFSerializerPackHeader)) {
+        snprintf(it->error_info, sizeof(it->error_info),
+                "content length: %d is too small which < %d",
+                content->len, (int)sizeof(SFSerializerPackHeader));
+        return EINVAL;
+    }
+
+    header = (SFSerializerPackHeader *)content->str;
+    length = buff2int(header->length);
+    if (content->len != length + sizeof(SFSerializerPackHeader)) {
+        snprintf(it->error_info, sizeof(it->error_info),
+                "content length: %d != %d", content->len,
+                (int)(length + sizeof(SFSerializerPackHeader)));
+        return EINVAL;
+    }
+
+    calc_crc32 = CRC32(header + 1, length);
+    header_crc32 = buff2int(header->crc32);
+    if (header_crc32 != calc_crc32) {
+        snprintf(it->error_info, sizeof(it->error_info),
+                "header crc32: %d != calculated: %d",
+                header_crc32, calc_crc32);
+        return EINVAL;
+    }
+
+    it->p = (const char *)(header + 1);
+    it->end = content->str + content->len;
+    return 0;
+}
+
+static int check_field_type(SFSerializerIterator *it,
+        const int remain_len, const SFSerializerValueType type)
+{
+    if (!(type >= 0 && type < SF_SERIALIZER_VALUE_TYPE_COUNT)) {
+        snprintf(it->error_info, sizeof(it->error_info),
+                "fid: %d, unknown type: %d", it->field.fid, type);
+        return EINVAL;
+    }
+
+    if (remain_len < value_type_configs[type].min_size) {
+        snprintf(it->error_info, sizeof(it->error_info),
+                FIELD_ID_AND_TYPE_FORMAT", remain length: %d "
+                "is too small which < %d", FIELD_ID_AND_TYPE_PARAMS,
+                remain_len, value_type_configs[type].min_size);
+        return EINVAL;
+    }
+    return 0;
+}
+
+static inline int check_string_value(SFSerializerIterator *it,
+        const int remain_len, const string_t *s)
+{
+    if (s->len < 0) {
+        snprintf(it->error_info, sizeof(it->error_info),
+                FIELD_ID_AND_TYPE_FORMAT", invalid string length: %d < 0",
+                FIELD_ID_AND_TYPE_PARAMS, s->len);
+        return EINVAL;
+    }
+
+    if (s->len > remain_len) {
+        snprintf(it->error_info, sizeof(it->error_info),
+                FIELD_ID_AND_TYPE_FORMAT", string length: %d is too "
+                "large > remain length: %d", FIELD_ID_AND_TYPE_PARAMS,
+                s->len, remain_len);
+        return EINVAL;
+    }
+
+    return 0;
+}
+
+static inline int unpack_array_count(SFSerializerIterator *it,
+        const int remain_len, int *count)
+{
+    int min_size;
+
+    *count = buff2int(((SFSerializerPackFieldArray *)it->p)->value.count);
+    if (*count < 0) {
+        snprintf(it->error_info, sizeof(it->error_info),
+                FIELD_ID_AND_TYPE_FORMAT", invalid array count: %d < 0",
+                FIELD_ID_AND_TYPE_PARAMS, *count);
+        return EINVAL;
+    }
+
+    min_size = value_type_configs[it->field.type].elt_size * (*count);
+    if (remain_len < min_size) {
+        snprintf(it->error_info, sizeof(it->error_info),
+                FIELD_ID_AND_TYPE_FORMAT", remain length: %d is too "
+                "small < array min bytes: %d", FIELD_ID_AND_TYPE_PARAMS,
+                remain_len, min_size);
+        return EINVAL;
+    }
+
+    return 0;
+}
+
+static int array_expand(SFSerializerIterator *it, void_array_t *array,
+        const int elt_size, const int target_count, int *alloc_size)
+{
+    int new_alloc;
+    void *new_elts;
+
+    if (*alloc_size == 0) {
+        new_alloc = 256;
+    } else {
+        new_alloc = (*alloc_size) * 2;
+    }
+    while (new_alloc < target_count) {
+        new_alloc *= 2;
+    }
+
+    new_elts = fc_malloc(elt_size * new_alloc);
+    if (new_elts == NULL) {
+        snprintf(it->error_info, sizeof(it->error_info),
+                FIELD_ID_AND_TYPE_FORMAT", malloc %d bytes fail",
+                FIELD_ID_AND_TYPE_PARAMS, elt_size * new_alloc);
+        return ENOMEM;
+    }
+
+    if (array->elts != NULL) {
+        free(array->elts);
+    }
+    array->elts = new_elts;
+    *alloc_size = new_alloc;
+    return 0;
+}
+
+static inline int unpack_string(SFSerializerIterator *it, const int remain_len,
+        SFSerializerPackStringValue *input, string_t *output)
+{
+    if (remain_len < sizeof(SFSerializerPackStringValue)) {
+        snprintf(it->error_info, sizeof(it->error_info),
+                FIELD_ID_AND_TYPE_FORMAT", remain length: %d "
+                "is too small < %d", FIELD_ID_AND_TYPE_PARAMS,
+                remain_len, (int)sizeof(SFSerializerPackStringValue));
+        return EINVAL;
+    }
+
+    output->len = buff2int(input->len);
+    output->str = input->str;
+    it->p += sizeof(SFSerializerPackStringValue) + output->len;
+    return check_string_value(it, remain_len -
+            sizeof(SFSerializerPackStringValue), output);
+}
+
+static int unpack_integer_array(SFSerializerIterator *it, const int remain_len)
+{
+    int result;
+    int count;
+    int64_t *pn;
+    int64_t *end;
+
+    if ((result=unpack_array_count(it, remain_len, &count)) != 0) {
+        return result;
+    }
+
+    if (count > it->int_array_alloc) {
+        if ((result=array_expand(it, (void_array_t *)&it->int_array,
+                        sizeof(int64_t), count, &it->int_array_alloc)) != 0)
+        {
+            return result;
+        }
+    }
+
+    it->p += sizeof(SFSerializerPackFieldArray);
+    end = it->int_array.elts + count;
+    for (pn=it->int_array.elts; pn<end; pn++) {
+        switch (it->field.type) {
+            case sf_serializer_value_type_int8_array:
+                *pn = *it->p;
+                break;
+            case sf_serializer_value_type_int16_array:
+                *pn = buff2short(it->p);
+                break;
+            case sf_serializer_value_type_int32_array:
+                *pn = buff2int(it->p);
+                break;
+            default:
+                *pn = buff2long(it->p);
+                break;
+        }
+        it->p += value_type_configs[it->field.type].elt_size;
+    }
+    it->int_array.count = count;
+
+    return 0;
+}
+
+static int unpack_string_array(SFSerializerIterator *it, const int remain_len)
+{
+    int result;
+    int count;
+    string_t *str;
+    string_t *end;
+
+    if ((result=unpack_array_count(it, remain_len, &count)) != 0) {
+        return result;
+    }
+
+    if (count > it->str_array_alloc) {
+        if ((result=array_expand(it, (void_array_t *)&it->str_array,
+                        sizeof(string_t), count, &it->str_array_alloc)) != 0)
+        {
+            return result;
+        }
+    }
+
+    it->p += sizeof(SFSerializerPackFieldArray);
+    end = it->str_array.strings + count;
+    for (str=it->str_array.strings; str<end; str++) {
+        if ((result=unpack_string(it, it->end - it->p,
+                        (SFSerializerPackStringValue *)
+                        it->p, str)) != 0)
+        {
+            return result;
+        }
+    }
+    it->str_array.count = count;
+
+    return 0;
+}
+
+static int unpack_id_name_array(SFSerializerIterator *it, const int remain_len)
+{
+    int result;
+    int count;
+    id_name_pair_t *pair;
+    id_name_pair_t *end;
+
+    if ((result=unpack_array_count(it, remain_len, &count)) != 0) {
+        return result;
+    }
+
+    if (count > it->id_name_array_alloc) {
+        if ((result=array_expand(it, (void_array_t *)&it->id_name_array,
+                        sizeof(id_name_pair_t), count,
+                        &it->id_name_array_alloc)) != 0)
+        {
+            return result;
+        }
+    }
+
+    it->p += sizeof(SFSerializerPackFieldArray);
+    end = it->id_name_array.elts + count;
+    for (pair=it->id_name_array.elts; pair<end; pair++) {
+        if ((it->end - it->p) < (sizeof(int64_t) +
+                    sizeof(SFSerializerPackStringValue)))
+        {
+            snprintf(it->error_info, sizeof(it->error_info),
+                    FIELD_ID_AND_TYPE_FORMAT", remain length: %d "
+                    "is too small < %d", FIELD_ID_AND_TYPE_PARAMS,
+                    (int)(it->end - it->p), (int)(sizeof(int64_t) +
+                        sizeof(SFSerializerPackStringValue)));
+            return EINVAL;
+        }
+
+        pair->id = buff2long(it->p);
+        it->p += sizeof(int64_t);
+        if ((result=unpack_string(it, it->end - it->p,
+                        (SFSerializerPackStringValue *)it->p,
+                        &pair->name)) != 0)
+        {
+            return result;
+        }
+    }
+    it->id_name_array.count = count;
+
+    return 0;
+}
+
+static int unpack_map(SFSerializerIterator *it, const int remain_len)
+{
+    int result;
+    int count;
+    key_value_pair_t *pair;
+    key_value_pair_t *end;
+
+    if ((result=unpack_array_count(it, remain_len, &count)) != 0) {
+        return result;
+    }
+
+    if (count > it->kv_array_alloc) {
+        if ((result=array_expand(it, (void_array_t *)&it->kv_array,
+                        sizeof(key_value_pair_t), count,
+                        &it->kv_array_alloc)) != 0)
+        {
+            return result;
+        }
+    }
+
+    it->p += sizeof(SFSerializerPackFieldArray);
+    end = it->kv_array.kv_pairs + count;
+    for (pair=it->kv_array.kv_pairs; pair<end; pair++) {
+        if ((result=unpack_string(it, it->end - it->p,
+                        (SFSerializerPackStringValue *)it->p,
+                        &pair->key)) != 0)
+        {
+            return result;
+        }
+        if ((result=unpack_string(it, it->end - it->p,
+                        (SFSerializerPackStringValue *)it->p,
+                        &pair->value)) != 0)
+        {
+            return result;
+        }
+    }
+    it->kv_array.count = count;
+
+    return 0;
+}
+
+const SFSerializerFieldValue *sf_serializer_next(SFSerializerIterator *it)
+{
+    int remain_len;
+    SFSerializerPackFieldInfo *field;
+    SFSerializerPackFieldString *fs;
+
+    remain_len = it->end - it->p;
+    if (remain_len == 0) {
+        return NULL;
+    }
+
+    if (remain_len <= sizeof(SFSerializerPackFieldInfo)) {
+        snprintf(it->error_info, sizeof(it->error_info),
+                "remain length: %d is too small which <= %d",
+                remain_len, (int)sizeof(SFSerializerPackFieldInfo));
+        it->error_no = EINVAL;
+        return NULL;
+    }
+
+    field = (SFSerializerPackFieldInfo *)it->p;
+    it->field.fid = field->id;
+    it->field.type = field->type;
+    if ((it->error_no=check_field_type(it, remain_len, field->type)) != 0) {
+        return NULL;
+    }
+
+    switch (field->type) {
+        case sf_serializer_value_type_int8:
+            it->field.value.n = ((SFSerializerPackFieldInt8 *)it->p)->value;
+            it->p += sizeof(SFSerializerPackFieldInt8);
+            break;
+        case sf_serializer_value_type_int16:
+            it->field.value.n = buff2short(
+                    ((SFSerializerPackFieldInt16 *)
+                     it->p)->value);
+            it->p += sizeof(SFSerializerPackFieldInt16);
+            break;
+        case sf_serializer_value_type_int32:
+            it->field.value.n = buff2int(
+                    ((SFSerializerPackFieldInt32 *)
+                     it->p)->value);
+            it->p += sizeof(SFSerializerPackFieldInt32);
+            break;
+        case sf_serializer_value_type_int64:
+            it->field.value.n = buff2long(
+                    ((SFSerializerPackFieldInt64 *)
+                     it->p)->value);
+            it->p += sizeof(SFSerializerPackFieldInt64);
+            break;
+        case sf_serializer_value_type_string:
+            fs = (SFSerializerPackFieldString *)it->p;
+            it->p += sizeof(SFSerializerPackFieldInfo);
+            if ((it->error_no=unpack_string(it, remain_len -
+                            sizeof(SFSerializerPackFieldInfo),
+                            &fs->value, &it->field.value.s)) != 0)
+            {
+                return NULL;
+            }
+            break;
+        case sf_serializer_value_type_int8_array:
+        case sf_serializer_value_type_int16_array:
+        case sf_serializer_value_type_int32_array:
+        case sf_serializer_value_type_int64_array:
+            if ((it->error_no=unpack_integer_array(it, remain_len -
+                            sizeof(SFSerializerPackFieldArray))) != 0)
+            {
+                return NULL;
+            }
+            it->field.value.int_array = it->int_array;
+            break;
+        case sf_serializer_value_type_string_array:
+            if ((it->error_no=unpack_string_array(it, remain_len - sizeof(
+                                SFSerializerPackFieldArray))) != 0)
+            {
+                return NULL;
+            }
+            it->field.value.str_array = it->str_array;
+            break;
+        case sf_serializer_value_type_id_name_array:
+            if ((it->error_no=unpack_id_name_array(it, remain_len -
+                            sizeof(SFSerializerPackFieldArray))) != 0)
+            {
+                return NULL;
+            }
+            it->field.value.id_name_array = it->id_name_array;
+            break;
+        case sf_serializer_value_type_map:
+            if ((it->error_no=unpack_map(it, remain_len - sizeof(
+                                SFSerializerPackFieldArray))) != 0)
+            {
+                return NULL;
+            }
+            it->field.value.kv_array = it->kv_array;
+            break;
+    }
+
+    return &it->field;
+}
+
+int sf_serializer_read_message(int fd, BufferInfo *buffer,
+        const int max_size)
+{
+    SFSerializerPackHeader *header;
+    char *new_buff;
+    int new_alloc;
+    int length;
+    int total_bytes;
+
+    if (fc_safe_read(fd, buffer->buff, sizeof(*header)) != sizeof(*header)) {
+        return ENODATA;
+    }
+
+    header = (SFSerializerPackHeader *)buffer->buff;
+    length = buff2int(header->length);
+    if (length <= 0 || length > max_size) {
+        return EINVAL;
+    }
+
+    total_bytes = sizeof(*header) + length;
+    if (buffer->alloc_size < total_bytes) {
+        new_alloc = buffer->alloc_size * 2;
+        while (new_alloc < total_bytes) {
+            new_alloc *= 2;
+        }
+
+        new_buff = (char *)fc_malloc(new_alloc);
+        if (new_buff == NULL) {
+            return ENOMEM;
+        }
+
+        memcpy(new_buff, buffer->buff, sizeof(*header));
+        free(buffer->buff);
+        buffer->buff = new_buff;
+        buffer->alloc_size = new_alloc;
+    }
+
+    if (fc_safe_read(fd, buffer->buff + sizeof(*header),
+                length) != length)
+    {
+        return ENODATA;
+    }
+
+    buffer->length = total_bytes;
+    return 0;
+}

src/sf_serializer.h

@@ -0,0 +1,563 @@
+/*
+ * Copyright (c) 2020 YuQing <384681@qq.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3
+ * or later ("AGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <https://www.gnu.org/licenses/>.
+ */
+
+//sf_serializer.h
+
+#ifndef _SF_SERIALIZER_H
+#define _SF_SERIALIZER_H
+
+#include "fastcommon/common_define.h"
+#include "fastcommon/shared_func.h"
+#include "fastcommon/fast_buffer.h"
+#include "fastcommon/uniq_skiplist.h"
+#include "fastcommon/hash.h"
+
+#define SF_SERIALIZER_VALUE_TYPE_COUNT  12
+
+typedef enum {
+    sf_serializer_value_type_int8 = 0,
+    sf_serializer_value_type_int16,
+    sf_serializer_value_type_int32,
+    sf_serializer_value_type_int64,
+    sf_serializer_value_type_string,
+    sf_serializer_value_type_int8_array,
+    sf_serializer_value_type_int16_array,
+    sf_serializer_value_type_int32_array,
+    sf_serializer_value_type_int64_array,
+    sf_serializer_value_type_string_array,
+    sf_serializer_value_type_id_name_array,
+    sf_serializer_value_type_map
+} SFSerializerValueType;
+
+typedef struct sf_serializer_pack_header {
+    char length[4];
+    char crc32[4];
+} SFSerializerPackHeader;
+
+typedef struct sf_serializer_pack_field_info {
+    unsigned char id;
+    unsigned char type;
+} SFSerializerPackFieldInfo;
+
+typedef struct sf_serializer_pack_field_int8 {
+    SFSerializerPackFieldInfo field;
+    char value;
+} SFSerializerPackFieldInt8;
+
+typedef struct sf_serializer_pack_field_int16 {
+    SFSerializerPackFieldInfo field;
+    char value[2];
+} SFSerializerPackFieldInt16;
+
+typedef struct sf_serializer_pack_field_int32 {
+    SFSerializerPackFieldInfo field;
+    char value[4];
+} SFSerializerPackFieldInt32;
+
+typedef struct sf_serializer_pack_field_int64 {
+    SFSerializerPackFieldInfo field;
+    char value[8];
+} SFSerializerPackFieldInt64;
+
+typedef struct sf_serializer_pack_string_value {
+    char len[4];
+    char str[0];
+} SFSerializerPackStringValue;
+
+typedef struct sf_serializer_pack_field_string {
+    SFSerializerPackFieldInfo field;
+    SFSerializerPackStringValue value;
+} SFSerializerPackFieldString;
+
+typedef struct sf_serializer_pack_field_array {
+    SFSerializerPackFieldInfo field;
+    struct {
+        char count[4];
+        char ptr[0];
+    } value;
+} SFSerializerPackFieldArray;
+
+typedef struct sf_serializer_field_value {
+    unsigned char fid;
+    SFSerializerValueType type;
+    union {
+        int64_t n;
+        string_t s;
+        int64_array_t int_array;
+        string_array_t str_array;
+        id_name_array_t id_name_array;
+        key_value_array_t kv_array;
+    } value;
+} SFSerializerFieldValue;
+
+typedef struct sf_serializer_iterator {
+    const char *p;
+    const char *end;
+    int64_array_t int_array;       //int64_t array holder
+    string_array_t str_array;      //string_t array holder
+    id_name_array_t id_name_array; //id name array holder
+    key_value_array_t kv_array;    //key value array holder
+    int int_array_alloc;
+    int str_array_alloc;
+    int id_name_array_alloc;
+    int kv_array_alloc;
+    SFSerializerFieldValue field;
+    int error_no;
+    char error_info[256];
+} SFSerializerIterator;
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+static inline void sf_serializer_pack_begin(FastBuffer *buffer)
+{
+    buffer->length = sizeof(SFSerializerPackHeader);
+}
+
+static inline int sf_serializer_pack_int8(FastBuffer *buffer,
+        const unsigned char fid, const int8_t value)
+{
+    int result;
+    SFSerializerPackFieldInt8 *obj;
+
+    if ((result=fast_buffer_check_inc_size(buffer,
+                    sizeof(SFSerializerPackFieldInt8))) != 0)
+    {
+        return result;
+    }
+
+    obj = (SFSerializerPackFieldInt8 *)(buffer->data + buffer->length);
+    obj->field.id = fid;
+    obj->field.type = sf_serializer_value_type_int8;
+    obj->value = value;
+    buffer->length += sizeof(SFSerializerPackFieldInt8);
+    return 0;
+}
+
+static inline int sf_serializer_pack_int16(FastBuffer *buffer,
+        const unsigned char fid, const int16_t value)
+{
+    int result;
+    SFSerializerPackFieldInt16 *obj;
+
+    if ((result=fast_buffer_check_inc_size(buffer,
+                    sizeof(SFSerializerPackFieldInt16))) != 0)
+    {
+        return result;
+    }
+
+    obj = (SFSerializerPackFieldInt16 *)(buffer->data + buffer->length);
+    obj->field.id = fid;
+    obj->field.type = sf_serializer_value_type_int16;
+    short2buff(value, obj->value);
+    buffer->length += sizeof(SFSerializerPackFieldInt16);
+    return 0;
+}
+
+static inline int sf_serializer_pack_int32(FastBuffer *buffer,
+        const unsigned char fid, const int32_t value)
+{
+    int result;
+    SFSerializerPackFieldInt32 *obj;
+
+    if ((result=fast_buffer_check_inc_size(buffer,
+                    sizeof(SFSerializerPackFieldInt32))) != 0)
+    {
+        return result;
+    }
+
+    obj = (SFSerializerPackFieldInt32 *)(buffer->data + buffer->length);
+    obj->field.id = fid;
+    obj->field.type = sf_serializer_value_type_int32;
+    int2buff(value, obj->value);
+    buffer->length += sizeof(SFSerializerPackFieldInt32);
+    return 0;
+}
+
+static inline int sf_serializer_pack_int64(FastBuffer *buffer,
+        const unsigned char fid, const int64_t value)
+{
+    int result;
+    SFSerializerPackFieldInt64 *obj;
+
+    if ((result=fast_buffer_check_inc_size(buffer,
+                    sizeof(SFSerializerPackFieldInt64))) != 0)
+    {
+        return result;
+    }
+
+    obj = (SFSerializerPackFieldInt64 *)(buffer->data + buffer->length);
+    obj->field.id = fid;
+    obj->field.type = sf_serializer_value_type_int64;
+    long2buff(value, obj->value);
+    buffer->length += sizeof(SFSerializerPackFieldInt64);
+    return 0;
+}
+
+static inline int sf_serializer_pack_integer(FastBuffer *buffer,
+        const unsigned char fid, const int64_t value)
+{
+    if (value >= INT16_MIN && value <= INT16_MAX) {
+        if (value >= INT8_MIN && value <= INT8_MAX) {
+            return sf_serializer_pack_int8(buffer, fid, value);
+        } else {
+            return sf_serializer_pack_int16(buffer, fid, value);
+        }
+    } else {
+        if (value >= INT32_MIN && value <= INT32_MAX) {
+            return sf_serializer_pack_int32(buffer, fid, value);
+        } else {
+            return sf_serializer_pack_int64(buffer, fid, value);
+        }
+    }
+}
+
+#define SF_SERIALIZER_PACK_STRING(ps, value)  \
+    int2buff((value)->len, (ps)->len);   \
+    memcpy((ps)->str, (value)->str, (value)->len)
+
+#define SF_SERIALIZER_PACK_STRING_AND_MOVE_PTR(p, value)  \
+    SF_SERIALIZER_PACK_STRING((SFSerializerPackStringValue *)p, value); \
+    p += (sizeof(SFSerializerPackStringValue) + (value)->len)
+
+static inline int sf_serializer_pack_string(FastBuffer *buffer,
+        const unsigned char fid, const string_t *value)
+{
+    int result;
+    int length;
+    SFSerializerPackFieldString *obj;
+
+    length = sizeof(SFSerializerPackFieldString) + value->len;
+    if ((result=fast_buffer_check_inc_size(buffer, length)) != 0) {
+        return result;
+    }
+
+    obj = (SFSerializerPackFieldString *)(buffer->data + buffer->length);
+    obj->field.id = fid;
+    obj->field.type = sf_serializer_value_type_string;
+    SF_SERIALIZER_PACK_STRING(&obj->value, value);
+    buffer->length += length;
+    return 0;
+}
+
+static inline int sf_serializer_pack_buffer(FastBuffer *buffer,
+        const unsigned char fid, const FastBuffer *value)
+{
+    string_t str;
+    FC_SET_STRING_EX(str, value->data, value->length);
+    return sf_serializer_pack_string(buffer, fid, &str);
+}
+
+static inline int sf_serializer_pack_int8_array(FastBuffer *buffer,
+        const unsigned char fid, const int8_t *array, const int count)
+{
+    int result;
+    int length;
+    SFSerializerPackFieldArray *obj;
+    const int8_t *pn;
+    const int8_t *end;
+    char *ps;
+
+    length = sizeof(SFSerializerPackFieldArray) + count * 1;
+    if ((result=fast_buffer_check_inc_size(buffer, length)) != 0) {
+        return result;
+    }
+
+    obj = (SFSerializerPackFieldArray *)(buffer->data + buffer->length);
+    obj->field.id = fid;
+    obj->field.type = sf_serializer_value_type_int8_array;
+    int2buff(count, obj->value.count);
+    end = array + count;
+    for (pn=array, ps=obj->value.ptr; pn<end; pn++) {
+        *ps++ = *pn;
+    }
+    buffer->length += length;
+    return 0;
+}
+
+static inline int sf_serializer_pack_int16_array(FastBuffer *buffer,
+        const unsigned char fid, const int16_t *array, const int count)
+{
+    int result;
+    int length;
+    SFSerializerPackFieldArray *obj;
+    const int16_t *pn;
+    const int16_t *end;
+    char *ps;
+
+    length = sizeof(SFSerializerPackFieldArray) + count * 2;
+    if ((result=fast_buffer_check_inc_size(buffer, length)) != 0) {
+        return result;
+    }
+
+    obj = (SFSerializerPackFieldArray *)(buffer->data + buffer->length);
+    obj->field.id = fid;
+    obj->field.type = sf_serializer_value_type_int16_array;
+    int2buff(count, obj->value.count);
+    end = array + count;
+    for (pn=array, ps=obj->value.ptr; pn<end; pn++, ps+=2) {
+        short2buff(*pn, ps);
+    }
+    buffer->length += length;
+    return 0;
+}
+
+static inline int sf_serializer_pack_int32_array(FastBuffer *buffer,
+        const unsigned char fid, const int32_t *array, const int count)
+{
+    int result;
+    int length;
+    SFSerializerPackFieldArray *obj;
+    const int32_t *pn;
+    const int32_t *end;
+    char *ps;
+
+    length = sizeof(SFSerializerPackFieldArray) + count * 4;
+    if ((result=fast_buffer_check_inc_size(buffer, length)) != 0) {
+        return result;
+    }
+
+    obj = (SFSerializerPackFieldArray *)(buffer->data + buffer->length);
+    obj->field.id = fid;
+    obj->field.type = sf_serializer_value_type_int32_array;
+    int2buff(count, obj->value.count);
+    end = array + count;
+    for (pn=array, ps=obj->value.ptr; pn<end; pn++, ps+=4) {
+        int2buff(*pn, ps);
+    }
+    buffer->length += length;
+    return 0;
+}
+
+static inline int sf_serializer_pack_int64_array(FastBuffer *buffer,
+        const unsigned char fid, const int64_t *array, const int count)
+{
+    int result;
+    int length;
+    SFSerializerPackFieldArray *obj;
+    const int64_t *pn;
+    const int64_t *end;
+    char *ps;
+
+    length = sizeof(SFSerializerPackFieldArray) + count * 8;
+    if ((result=fast_buffer_check_inc_size(buffer, length)) != 0) {
+        return result;
+    }
+
+    obj = (SFSerializerPackFieldArray *)(buffer->data + buffer->length);
+    obj->field.id = fid;
+    obj->field.type = sf_serializer_value_type_int64_array;
+    int2buff(count, obj->value.count);
+    end = array + count;
+    for (pn=array, ps=obj->value.ptr; pn<end; pn++, ps+=8) {
+        long2buff(*pn, ps);
+    }
+    buffer->length += length;
+    return 0;
+}
+
+static inline int sf_serializer_pack_string_array(FastBuffer *buffer,
+        const unsigned char fid, const string_t *strings, const int count)
+{
+    int result;
+    int length;
+    SFSerializerPackFieldArray *obj;
+    const string_t *str;
+    const string_t *end;
+    char *p;
+
+    length = sizeof(SFSerializerPackFieldArray);
+    end = strings + count;
+    for (str=strings; str<end; str++) {
+        length += sizeof(SFSerializerPackStringValue) + str->len;
+    }
+
+    if ((result=fast_buffer_check_inc_size(buffer, length)) != 0) {
+        return result;
+    }
+
+    obj = (SFSerializerPackFieldArray *)(buffer->data + buffer->length);
+    obj->field.id = fid;
+    obj->field.type = sf_serializer_value_type_string_array;
+    int2buff(count, obj->value.count);
+
+    p = obj->value.ptr;
+    for (str=strings; str<end; str++) {
+        SF_SERIALIZER_PACK_STRING_AND_MOVE_PTR(p, str);
+    }
+    buffer->length += length;
+    return 0;
+}
+
+static inline int sf_serializer_pack_id_name_array(FastBuffer *buffer,
+        const unsigned char fid, const id_name_pair_t *in_pairs,
+        const int count)
+{
+    int result;
+    int length;
+    SFSerializerPackFieldArray *obj;
+    const id_name_pair_t *pair;
+    const id_name_pair_t *end;
+    char *p;
+
+    length = sizeof(SFSerializerPackFieldArray);
+    end = in_pairs + count;
+    for (pair=in_pairs; pair<end; pair++) {
+        length += sizeof(int64_t) + pair->name.len +
+            sizeof(SFSerializerPackStringValue);
+    }
+
+    if ((result=fast_buffer_check_inc_size(buffer, length)) != 0) {
+        return result;
+    }
+
+    obj = (SFSerializerPackFieldArray *)(buffer->data + buffer->length);
+    obj->field.id = fid;
+    obj->field.type = sf_serializer_value_type_id_name_array;
+    int2buff(count, obj->value.count);
+
+    p = obj->value.ptr;
+    for (pair=in_pairs; pair<end; pair++) {
+        long2buff(pair->id, p);
+        p += sizeof(int64_t);
+        SF_SERIALIZER_PACK_STRING_AND_MOVE_PTR(p, &pair->name);
+    }
+    buffer->length += length;
+    return 0;
+}
+
+static inline int sf_serializer_pack_id_name_skiplist(
+        FastBuffer *buffer, const unsigned char fid,
+        UniqSkiplist *sl)
+{
+    int result;
+    int length;
+    SFSerializerPackFieldArray *obj;
+    const id_name_pair_t *pair;
+    UniqSkiplistIterator it;
+    char *p;
+
+    length = sizeof(SFSerializerPackFieldArray);
+    uniq_skiplist_iterator(sl, &it);
+    while ((pair=uniq_skiplist_next(&it)) != NULL) {
+        length += sizeof(int64_t) + pair->name.len +
+            sizeof(SFSerializerPackStringValue);
+    }
+
+    if ((result=fast_buffer_check_inc_size(buffer, length)) != 0) {
+        return result;
+    }
+
+    obj = (SFSerializerPackFieldArray *)(buffer->data + buffer->length);
+    obj->field.id = fid;
+    obj->field.type = sf_serializer_value_type_id_name_array;
+    int2buff(uniq_skiplist_count(sl), obj->value.count);
+
+    p = obj->value.ptr;
+    uniq_skiplist_iterator(sl, &it);
+    while ((pair=uniq_skiplist_next(&it)) != NULL) {
+        long2buff(pair->id, p);
+        p += sizeof(int64_t);
+        SF_SERIALIZER_PACK_STRING_AND_MOVE_PTR(p, &pair->name);
+    }
+    buffer->length += length;
+    return 0;
+}
+
+static inline int sf_serializer_pack_map(FastBuffer *buffer,
+        const unsigned char fid, const key_value_pair_t *kv_pairs,
+        const int count)
+{
+    int result;
+    int length;
+    SFSerializerPackFieldArray *obj;
+    const key_value_pair_t *pair;
+    const key_value_pair_t *end;
+    char *p;
+
+    length = sizeof(SFSerializerPackFieldArray);
+    end = kv_pairs + count;
+    for (pair=kv_pairs; pair<end; pair++) {
+        length += sizeof(SFSerializerPackStringValue) * 2 +
+            pair->key.len + pair->value.len;
+    }
+
+    if ((result=fast_buffer_check_inc_size(buffer, length)) != 0) {
+        return result;
+    }
+
+    obj = (SFSerializerPackFieldArray *)(buffer->data + buffer->length);
+    obj->field.id = fid;
+    obj->field.type = sf_serializer_value_type_map;
+    int2buff(count, obj->value.count);
+
+    p = obj->value.ptr;
+    for (pair=kv_pairs; pair<end; pair++) {
+        SF_SERIALIZER_PACK_STRING_AND_MOVE_PTR(p, &pair->key);
+        SF_SERIALIZER_PACK_STRING_AND_MOVE_PTR(p, &pair->value);
+    }
+    buffer->length += length;
+    return 0;
+}
+
+static inline void sf_serializer_pack_end(FastBuffer *buffer)
+{
+    SFSerializerPackHeader *header;
+    int length;
+    int crc32;
+
+    header = (SFSerializerPackHeader *)buffer->data;
+    length = buffer->length - sizeof(SFSerializerPackHeader);
+    crc32 = CRC32(header + 1, length);
+    int2buff(length, header->length);
+    int2buff(crc32, header->crc32);
+}
+
+static inline void sf_serializer_iterator_init(SFSerializerIterator *it)
+{
+    memset(it, 0, sizeof(SFSerializerIterator));
+}
+
+static inline void sf_serializer_iterator_destroy(SFSerializerIterator *it)
+{
+    if (it->int_array.elts != NULL) {
+        free(it->int_array.elts);
+        it->int_array_alloc = 0;
+    }
+
+    if (it->kv_array.kv_pairs != NULL) {
+        free(it->kv_array.kv_pairs);
+        it->kv_array_alloc = 0;
+    }
+
+    if (it->str_array.strings != NULL) {
+        free(it->str_array.strings);
+        it->str_array_alloc = 0;
+    }
+}
+
+int sf_serializer_unpack(SFSerializerIterator *it, const string_t *content);
+
+const SFSerializerFieldValue *sf_serializer_next(SFSerializerIterator *it);
+
+int sf_serializer_read_message(int fd, BufferInfo *buffer,
+        const int max_size);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif

src/sf_service.c

@@ -25,6 +25,7 @@
 #include <string.h>
 #include <errno.h>
 #include <fcntl.h>
+#include <ifaddrs.h>
 #include "fastcommon/logger.h"
 #include "fastcommon/sockopt.h"
 #include "fastcommon/shared_func.h"
@@ -32,8 +33,8 @@
 #include "fastcommon/sched_thread.h"
 #include "fastcommon/ioevent_loop.h"
 #include "fastcommon/fc_memory.h"
-#include "sf_global.h"
-#include "sf_nio.h"
+#include "sf_proto.h"
+#include "sf_util.h"
 #include "sf_service.h"
 
 #if defined(OS_LINUX)
@@ -56,91 +57,96 @@ struct worker_thread_context {
     struct nio_thread_data *thread_data;
 };
 
-struct accept_thread_context {
-    SFContext *sf_context;
-    int server_sock;
-};
-
-
-int sf_init_task(struct fast_task_info *task)
-{
-    task->connect_timeout = SF_G_CONNECT_TIMEOUT; //for client side
-    task->network_timeout = SF_G_NETWORK_TIMEOUT;
-    return 0;
-}
-
 static void *worker_thread_entrance(void *arg);
 
-static int sf_init_free_queues(const int task_arg_size,
-        TaskInitCallback init_callback)
+static int sf_init_free_queue(SFContext *sf_context, const char *name,
+        const bool double_buffers, const bool need_shrink_task_buffer,
+        const int task_padding_size, const int task_arg_size,
+        TaskInitCallback init_callback, void *init_arg)
 {
-#define ALLOC_CONNECTIONS_ONCE 1024
-
-    static bool sf_inited = false;
     int result;
+    int buffer_size;
     int m;
-    int init_connections;
+    int max_m;
     int alloc_conn_once;
 
-    if (sf_inited) {
-        return 0;
-    }
-
-    sf_inited = true;
     if ((result=set_rand_seed()) != 0) {
         logCrit("file: "__FILE__", line: %d, "
                 "set_rand_seed fail, program exit!", __LINE__);
         return result;
     }
 
-    m = g_sf_global_vars.min_buff_size / (64 * 1024);
+    if (strcmp(name, "cluster") == 0 || strcmp(name, "replica") == 0) {
+        buffer_size = FC_MAX(4 * 1024 * 1024, sf_context->
+                net_buffer_cfg.max_buff_size);
+        max_m = 64;
+    } else {
+        buffer_size = sf_context->net_buffer_cfg.min_buff_size;
+        max_m = 16;
+    }
+    m = buffer_size / (64 * 1024);
     if (m == 0) {
         m = 1;
-    } else if (m > 16) {
-        m = 16;
-    }
-    alloc_conn_once = ALLOC_CONNECTIONS_ONCE / m;
-    init_connections = g_sf_global_vars.max_connections < alloc_conn_once ?
-        g_sf_global_vars.max_connections : alloc_conn_once;
-    if ((result=free_queue_init_ex2(g_sf_global_vars.max_connections,
-                    init_connections, alloc_conn_once, g_sf_global_vars.
-                    min_buff_size, g_sf_global_vars.max_buff_size,
-                    task_arg_size, init_callback != NULL ?
-                    init_callback : sf_init_task)) != 0)
-    {
-        return result;
+    } else if (m > max_m) {
+        m = max_m;
     }
+    alloc_conn_once = 256 / m;
 
-    return 0;
+    return free_queue_init_ex2(&sf_context->free_queue, name, double_buffers,
+            need_shrink_task_buffer, sf_context->net_buffer_cfg.max_connections,
+            alloc_conn_once, sf_context->net_buffer_cfg.min_buff_size,
+            sf_context->net_buffer_cfg.max_buff_size, task_padding_size,
+            task_arg_size, init_callback, init_arg);
 }
 
-int sf_service_init_ex2(SFContext *sf_context,
+int sf_service_init_ex2(SFContext *sf_context, const char *name,
         sf_alloc_thread_extra_data_callback
         alloc_thread_extra_data_callback,
         ThreadLoopCallback thread_loop_callback,
         sf_accept_done_callback accept_done_callback,
         sf_set_body_length_callback set_body_length_func,
-        sf_deal_task_func deal_func, TaskCleanUpCallback task_cleanup_func,
+        sf_alloc_recv_buffer_callback alloc_recv_buffer_func,
+        sf_send_done_callback send_done_callback,
+        sf_deal_task_callback deal_func, TaskCleanUpCallback task_cleanup_func,
         sf_recv_timeout_callback timeout_callback, const int net_timeout_ms,
-        const int proto_header_size, const int task_arg_size,
-        TaskInitCallback init_callback)
+        const int proto_header_size, const int task_padding_size,
+        const int task_arg_size, const bool double_buffers,
+        const bool need_shrink_task_buffer, const bool explicit_post_recv,
+        TaskInitCallback init_callback, void *init_arg,
+        sf_release_buffer_callback release_buffer_callback)
 {
     int result;
     int bytes;
+    int extra_events;
+    int max_entries;
+    int i;
     struct worker_thread_context *thread_contexts;
     struct worker_thread_context *thread_ctx;
     struct nio_thread_data *thread_data;
-    struct nio_thread_data *pDataEnd;
+    struct nio_thread_data *data_end;
     pthread_t tid;
     pthread_attr_t thread_attr;
 
-    sf_context->realloc_task_buffer = g_sf_global_vars.
-                    min_buff_size < g_sf_global_vars.max_buff_size;
-    sf_context->accept_done_func = accept_done_callback;
-    sf_set_parameters_ex(sf_context, proto_header_size, set_body_length_func,
-            deal_func, task_cleanup_func, timeout_callback);
+    fc_safe_strcpy(sf_context->name, name);
+    sf_context->connect_need_log = true;
+    sf_context->realloc_task_buffer = sf_context->net_buffer_cfg.
+        min_buff_size < sf_context->net_buffer_cfg.max_buff_size;
+    sf_context->callbacks.accept_done = accept_done_callback;
+    sf_set_parameters_ex(sf_context, proto_header_size,
+            set_body_length_func, alloc_recv_buffer_func,
+            send_done_callback, deal_func, task_cleanup_func,
+            timeout_callback, release_buffer_callback);
+    if (explicit_post_recv) {
+        for (i=0; i<SF_ADDRESS_FAMILY_COUNT; i++) {
+            sf_context->handlers[i].handlers[SF_RDMACM_NETWORK_HANDLER_INDEX].
+                explicit_post_recv = true;
+        }
+    }
 
-    if ((result=sf_init_free_queues(task_arg_size, init_callback)) != 0) {
+    if ((result=sf_init_free_queue(sf_context, name, double_buffers,
+                    need_shrink_task_buffer, task_padding_size,
+                    task_arg_size, init_callback, init_arg)) != 0)
+    {
         return result;
     }
 
@@ -165,11 +171,71 @@ int sf_service_init_ex2(SFContext *sf_context,
         return ENOMEM;
     }
 
-    sf_context->thread_count = 0;
-    pDataEnd = sf_context->thread_data + sf_context->work_threads;
-    for (thread_data=sf_context->thread_data,thread_ctx=thread_contexts;
-            thread_data<pDataEnd; thread_data++,thread_ctx++)
+    if (SF_G_EPOLL_EDGE_TRIGGER) {
+#ifdef OS_LINUX
+#if IOEVENT_USE_EPOLL
+        extra_events = EPOLLET;
+#else
+        extra_events = 0;
+#endif
+#elif defined(OS_FREEBSD)
+        extra_events = EV_CLEAR;
+#else
+        extra_events = 0;
+#endif
+    } else {
+        extra_events = 0;
+    }
+
+    max_entries = (sf_context->net_buffer_cfg.max_connections +
+            sf_context->work_threads - 1) / sf_context->work_threads;
+    if (strcmp(sf_context->name, "cluster") == 0 ||
+            strcmp(sf_context->name, "replica") == 0)
     {
+        if (max_entries < 1024) {
+            max_entries += 8;
+        } else {
+            max_entries = 1024;
+        }
+    } else {
+        if (max_entries < 4 * 1024) {
+            max_entries = max_entries * 2;
+        } else if (max_entries < 8 * 1024) {
+            max_entries = (max_entries * 3) / 2;
+        } else if (max_entries < 16 * 1024) {
+            max_entries = (max_entries * 5) / 4;
+        } else if (max_entries < 32 * 1024) {
+            max_entries = (max_entries * 6) / 5;
+        } else if (max_entries < 64 * 1024) {
+            max_entries = (max_entries * 11) / 10;
+        } else if (max_entries < 128 * 1024) {
+            max_entries = (max_entries * 21) / 20;
+        }
+
+#if IOEVENT_USE_URING
+        if (sf_context->use_io_uring) {
+            if (max_entries > 32 * 1024) {
+                max_entries = 32 * 1024;
+            }
+        }
+#endif
+    }
+
+    g_current_time = time(NULL);
+    sf_context->thread_count = 0;
+    data_end = sf_context->thread_data + sf_context->work_threads;
+    for (thread_data=sf_context->thread_data,thread_ctx=thread_contexts;
+            thread_data<data_end; thread_data++,thread_ctx++)
+    {
+        thread_data->timeout_ms = net_timeout_ms;
+        FC_INIT_LIST_HEAD(&thread_data->polling_queue);
+        if (sf_context->smart_polling.enabled) {
+            thread_data->busy_polling_callback =
+                sf_rdma_busy_polling_callback;
+        } else {
+            thread_data->busy_polling_callback = NULL;
+        }
+
         thread_data->thread_loop_callback = thread_loop_callback;
         if (alloc_thread_extra_data_callback != NULL) {
             thread_data->arg = alloc_thread_extra_data_callback(
@@ -179,32 +245,60 @@ int sf_service_init_ex2(SFContext *sf_context,
             thread_data->arg = NULL;
         }
 
-        if (ioevent_init(&thread_data->ev_puller,
-            g_sf_global_vars.max_connections + 2, net_timeout_ms, 0) != 0)
+        if ((result=ioevent_init(&thread_data->ev_puller, sf_context->
+                        name, sf_context->use_io_uring, max_entries,
+                        net_timeout_ms, extra_events)) != 0)
         {
-            result  = errno != 0 ? errno : ENOMEM;
+            char prompt[256];
+#if IOEVENT_USE_URING
+            if (sf_context->use_io_uring) {
+                if (result == EPERM) {
+                    strcpy(prompt, " make sure kernel."
+                            "io_uring_disabled set to 0");
+                } else if (result == EINVAL) {
+                    sprintf(prompt, " maybe max_connections: %d is too large"
+                            " or [%s]'s work_threads: %d is too small",
+                            sf_context->net_buffer_cfg.max_connections,
+                            sf_context->name, sf_context->work_threads);
+                } else {
+                    *prompt = '\0';
+                }
+            } else {
+#endif
+                *prompt = '\0';
+#if IOEVENT_USE_URING
+            }
+#endif
+
             logError("file: "__FILE__", line: %d, "
-                "ioevent_init fail, "
-                "errno: %d, error info: %s",
-                __LINE__, result, strerror(result));
+                "ioevent_init fail, errno: %d, error info: %s.%s"
+                , __LINE__, result, strerror(result), prompt);
             return result;
         }
 
-        result = fast_timer_init(&thread_data->timer,
-                2 * g_sf_global_vars.network_timeout, g_current_time);
+#if IOEVENT_USE_URING
+        if (sf_context->use_io_uring && send_done_callback != NULL) {
+            ioevent_set_send_zc_done_notify(&thread_data->ev_puller, true);
+        }
+#endif
+
+        result = fast_timer_init(&thread_data->timer, 2 * sf_context->
+                net_buffer_cfg.network_timeout, g_current_time);
         if (result != 0) {
             logError("file: "__FILE__", line: %d, "
-                "fast_timer_init fail, "
-                "errno: %d, error info: %s",
-                __LINE__, result, strerror(result));
+                    "fast_timer_init fail, errno: %d, error info: %s",
+                    __LINE__, result, strerror(result));
             return result;
         }
 
-        if ((result=init_pthread_lock(&thread_data->waiting_queue.lock)) != 0) {
+        if ((result=init_pthread_lock(&thread_data->
+                        waiting_queue.lock)) != 0)
+        {
             return result;
         }
 #if defined(OS_LINUX)
-        FC_NOTIFY_READ_FD(thread_data) = eventfd(0, EFD_NONBLOCK);
+        FC_NOTIFY_READ_FD(thread_data) = eventfd(0,
+                EFD_NONBLOCK | EFD_CLOEXEC);
         if (FC_NOTIFY_READ_FD(thread_data) < 0) {
             result = errno != 0 ? errno : EPERM;
             logError("file: "__FILE__", line: %d, "
@@ -228,12 +322,14 @@ int sf_service_init_ex2(SFContext *sf_context,
         {
             break;
         }
+        FC_SET_CLOEXEC(FC_NOTIFY_READ_FD(thread_data));
+        FC_SET_CLOEXEC(FC_NOTIFY_WRITE_FD(thread_data));
 #endif
 
         thread_ctx->sf_context = sf_context;
         thread_ctx->thread_data = thread_data;
         if ((result=pthread_create(&tid, &thread_attr,
-            worker_thread_entrance, thread_ctx)) != 0)
+                        worker_thread_entrance, thread_ctx)) != 0)
         {
             logError("file: "__FILE__", line: %d, "
                     "create thread failed, startup threads: %d, "
@@ -250,11 +346,11 @@ int sf_service_init_ex2(SFContext *sf_context,
 
 int sf_service_destroy_ex(SFContext *sf_context)
 {
-    struct nio_thread_data *pDataEnd, *thread_data;
+    struct nio_thread_data *data_end, *thread_data;
 
-    free_queue_destroy();
-    pDataEnd = sf_context->thread_data + sf_context->work_threads;
-    for (thread_data=sf_context->thread_data; thread_data<pDataEnd;
+    free_queue_destroy(&sf_context->free_queue);
+    data_end = sf_context->thread_data + sf_context->work_threads;
+    for (thread_data=sf_context->thread_data; thread_data<data_end;
             thread_data++)
     {
         fast_timer_destroy(&thread_data->timer);
@@ -264,12 +360,36 @@ int sf_service_destroy_ex(SFContext *sf_context)
     return 0;
 }
 
+void sf_service_set_thread_loop_callback_ex(SFContext *sf_context,
+        ThreadLoopCallback thread_loop_callback)
+{
+    struct nio_thread_data *data_end, *thread_data;
+
+    data_end = sf_context->thread_data + sf_context->work_threads;
+    for (thread_data=sf_context->thread_data; thread_data<data_end;
+            thread_data++)
+    {
+        thread_data->thread_loop_callback = thread_loop_callback;
+    }
+}
+
 static void *worker_thread_entrance(void *arg)
 {
     struct worker_thread_context *thread_ctx;
     int thread_count;
 
     thread_ctx = (struct worker_thread_context *)arg;
+
+#ifdef OS_LINUX
+    {
+        char thread_name[32];
+        snprintf(thread_name, sizeof(thread_name), "%s-net[%d]",
+                thread_ctx->sf_context->name, (int)(thread_ctx->
+                    thread_data - thread_ctx->sf_context->thread_data));
+        prctl(PR_SET_NAME, thread_name);
+    }
+#endif
+
     thread_count = __sync_add_and_fetch(&thread_ctx->
             sf_context->thread_count, 1);
 
@@ -281,7 +401,7 @@ static void *worker_thread_entrance(void *arg)
 
     ioevent_loop(thread_ctx->thread_data,
             sf_recv_notify_read,
-            thread_ctx->sf_context->task_cleanup_func,
+            thread_ctx->sf_context->callbacks.task_cleanup,
             &g_sf_global_vars.continue_flag);
     ioevent_destroy(&thread_ctx->thread_data->ev_puller);
 
@@ -296,15 +416,20 @@ static void *worker_thread_entrance(void *arg)
     return NULL;
 }
 
-static int _socket_server(const char *bind_addr, int port, int *sock)
+int sf_socket_create_server(SFListener *listener,
+        int af, const char *bind_addr)
 {
     int result;
-    *sock = socketServer(bind_addr, port, &result);
-    if (*sock < 0) {
+
+    listener->sock = socketServer2(af, bind_addr,
+            listener->port, &result);
+    if (listener->sock < 0) {
         return result;
     }
 
-    if ((result=tcpsetserveropt(*sock, g_sf_global_vars.network_timeout)) != 0) {
+    if ((result=tcpsetserveropt(listener->sock, listener->handler->
+                    fh->ctx->net_buffer_cfg.network_timeout)) != 0)
+    {
         return result;
     }
 
@@ -314,101 +439,215 @@ static int _socket_server(const char *bind_addr, int port, int *sock)
 int sf_socket_server_ex(SFContext *sf_context)
 {
     int result;
+    int i;
+    bool dual_ports;
     const char *bind_addr;
+    SFAddressFamilyHandler *fh;
+    SFNetworkHandler *handler;
+    SFNetworkHandler *end;
 
-    sf_context->inner_sock = sf_context->outer_sock = -1;
-    if (sf_context->outer_port == sf_context->inner_port) {
-        if (*sf_context->outer_bind_addr == '\0' ||
-                *sf_context->inner_bind_addr == '\0') {
-            bind_addr = "";
-            return _socket_server(bind_addr, sf_context->outer_port,
-                    &sf_context->outer_sock);
-        } else if (strcmp(sf_context->outer_bind_addr,
-                    sf_context->inner_bind_addr) == 0) {
-            bind_addr = sf_context->outer_bind_addr;
-            if (is_private_ip(bind_addr)) {
-                return _socket_server(bind_addr, sf_context->
-                        inner_port, &sf_context->inner_sock);
-            } else {
-                return _socket_server(bind_addr, sf_context->
-                        outer_port, &sf_context->outer_sock);
-            }
+    for (i=0; i<SF_ADDRESS_FAMILY_COUNT; i++) {
+        fh = sf_context->handlers + i;
+        if (fh->af == AF_UNSPEC) {
+            continue;
         }
-    }
 
-    if ((result=_socket_server(sf_context->outer_bind_addr,
-                    sf_context->outer_port, &sf_context->outer_sock)) != 0)
-    {
-        return result;
-    }
+        end = fh->handlers + SF_NETWORK_HANDLER_COUNT;
+        for (handler=fh->handlers; handler<end; handler++) {
+            if (!handler->enabled) {
+                continue;
+            }
 
-    if ((result=_socket_server(sf_context->inner_bind_addr,
-                    sf_context->inner_port, &sf_context->inner_sock)) != 0)
-    {
-        return result;
+            handler->inner.enabled = false;
+            handler->outer.enabled = false;
+            if (handler->outer.port == handler->inner.port) {
+                if (*fh->outer_bind_addr == '\0' ||
+                        *fh->inner_bind_addr == '\0')
+                {
+                    bind_addr = "";
+                    if ((result=handler->create_server(&handler->
+                                    outer, fh->af, bind_addr)) != 0)
+                    {
+                        return result;
+                    }
+                    handler->outer.enabled = true;
+                    dual_ports = false;
+                } else if (strcmp(fh->outer_bind_addr,
+                            fh->inner_bind_addr) == 0)
+                {
+                    bind_addr = fh->outer_bind_addr;
+                    if (is_private_ip(bind_addr)) {
+                        if ((result=handler->create_server(&handler->
+                                        inner, fh->af, bind_addr)) != 0)
+                        {
+                            return result;
+                        }
+                        handler->inner.enabled = true;
+                    } else {
+                        if ((result=handler->create_server(&handler->
+                                        outer, fh->af, bind_addr)) != 0)
+                        {
+                            return result;
+                        }
+                        handler->outer.enabled = true;
+                    }
+                    dual_ports = false;
+                } else {
+                    dual_ports = true;
+                }
+            } else {
+                dual_ports = true;
+            }
+
+            if (dual_ports) {
+                if ((result=handler->create_server(&handler->outer,
+                                fh->af, fh->outer_bind_addr)) != 0)
+                {
+                    return result;
+                }
+
+                if ((result=handler->create_server(&handler->inner,
+                                fh->af, fh->inner_bind_addr)) != 0)
+                {
+                    return result;
+                }
+                handler->inner.enabled = true;
+                handler->outer.enabled = true;
+            }
+
+            /*
+               logInfo("%p [%d] inner {port: %d, enabled: %d}, "
+               "outer {port: %d, enabled: %d}", sf_context,
+               (int)(handler-sf_context->handlers),
+               handler->inner.port, handler->inner.enabled,
+               handler->outer.port, handler->outer.enabled);
+             */
+        }
     }
 
     return 0;
 }
 
-static void *accept_thread_entrance(void *arg)
+void sf_socket_close_server(SFListener *listener)
+{
+    if (listener->sock >= 0) {
+        close(listener->sock);
+        listener->sock = -1;
+    }
+}
+
+struct fast_task_info *sf_socket_accept_connection(SFListener *listener)
 {
-    struct accept_thread_context *accept_context;
     int incomesock;
     int port;
-    struct sockaddr_in inaddr;
     socklen_t sockaddr_len;
     struct fast_task_info *task;
 
-    accept_context = (struct accept_thread_context *)arg;
-    while (g_sf_global_vars.continue_flag) {
-        sockaddr_len = sizeof(inaddr);
-        incomesock = accept(accept_context->server_sock,
-                (struct sockaddr*)&inaddr, &sockaddr_len);
-        if (incomesock < 0) { //error
-            if (!(errno == EINTR || errno == EAGAIN)) {
-                logError("file: "__FILE__", line: %d, "
-                        "accept fail, errno: %d, error info: %s",
-                        __LINE__, errno, strerror(errno));
+    sockaddr_len = sizeof(listener->inaddr);
+    incomesock = accept(listener->sock, (struct sockaddr *)
+            &listener->inaddr, &sockaddr_len);
+    if (incomesock < 0) { //error
+        if (!(errno == EINTR || errno == EAGAIN)) {
+            logError("file: "__FILE__", line: %d, "
+                    "accept fail, errno: %d, error info: %s",
+                    __LINE__, errno, strerror(errno));
+        }
+
+        return NULL;
+    }
+
+    if (tcpsetnonblockopt(incomesock) != 0) {
+        close(incomesock);
+        return NULL;
+    }
+    FC_SET_CLOEXEC(incomesock);
+
+    if ((task=sf_alloc_init_server_task(listener->handler,
+                    incomesock)) == NULL)
+    {
+        close(incomesock);
+        return NULL;
+    }
+
+    getPeerIpAddPort(incomesock, task->client_ip,
+            sizeof(task->client_ip), &port);
+    task->port = port;
+    return task;
+}
+
+void sf_socket_close_ex(SFContext *sf_context)
+{
+    int i;
+    SFNetworkHandler *handler;
+    SFNetworkHandler *end;
+
+    for (i=0; i<SF_ADDRESS_FAMILY_COUNT; i++) {
+        if (sf_context->handlers[i].af == AF_UNSPEC) {
+            continue;
+        }
+
+        end = sf_context->handlers[i].handlers + SF_NETWORK_HANDLER_COUNT;
+        for (handler=sf_context->handlers[i].handlers; handler<end; handler++) {
+            if (!handler->enabled) {
+                continue;
             }
 
+            if (handler->outer.enabled) {
+                handler->close_server(&handler->outer);
+            }
+            if (handler->inner.enabled) {
+                handler->close_server(&handler->inner);
+            }
+        }
+    }
+}
+
+static void accept_run(SFListener *listener)
+{
+    struct fast_task_info *task;
+
+    while (g_sf_global_vars.continue_flag) {
+        if ((task=listener->handler->accept_connection(listener)) == NULL) {
             continue;
         }
 
-        if (tcpsetnonblockopt(incomesock) != 0) {
-            close(incomesock);
-            continue;
-        }
-
-        if ((task=sf_alloc_init_task(accept_context->
-                        sf_context, incomesock)) == NULL)
-        {
-            close(incomesock);
-            continue;
-        }
-
-        getPeerIpAddPort(incomesock, task->client_ip,
-                sizeof(task->client_ip), &port);
-        task->port = port;
-        task->thread_data = accept_context->sf_context->thread_data +
-            incomesock % accept_context->sf_context->work_threads;
-        if (accept_context->sf_context->accept_done_func != NULL) {
-            accept_context->sf_context->accept_done_func(task,
-                    accept_context->server_sock ==
-                    accept_context->sf_context->inner_sock);
+        task->thread_data = listener->handler->fh->ctx->thread_data +
+            task->event.fd % listener->handler->fh->ctx->work_threads;
+        if (listener->handler->fh->ctx->callbacks.accept_done != NULL) {
+            if (listener->handler->fh->ctx->callbacks.accept_done(task,
+                        listener->inaddr.sin_addr.s_addr,
+                        listener->is_inner) != 0)
+            {
+                listener->handler->close_connection(task);
+                sf_release_task(task);
+                continue;
+            }
         }
 
         if (sf_nio_notify(task, SF_NIO_STAGE_INIT) != 0) {
-            close(incomesock);
+            listener->handler->close_connection(task);
             sf_release_task(task);
         }
     }
+}
 
+static void *accept_thread_entrance(SFListener *listener)
+{
+#ifdef OS_LINUX
+    {
+        char thread_name[32];
+        snprintf(thread_name, sizeof(thread_name), "%s-%s-listen",
+                listener->handler->comm_type == fc_comm_type_sock ?
+                "sock" : "rdma", listener->handler->fh->ctx->name);
+        prctl(PR_SET_NAME, thread_name);
+    }
+#endif
+
+    accept_run(listener);
     return NULL;
 }
 
-void _accept_loop(struct accept_thread_context *accept_context,
-        const int accept_threads)
+int _accept_loop(SFListener *listener, const int accept_threads)
 {
     pthread_t tid;
     pthread_attr_t thread_attr;
@@ -416,7 +655,7 @@ void _accept_loop(struct accept_thread_context *accept_context,
     int i;
 
     if (accept_threads <= 0) {
-       return;
+        return 0;
     }
 
     if ((result=init_pthread_attr(&thread_attr, g_sf_global_vars.
@@ -424,68 +663,83 @@ void _accept_loop(struct accept_thread_context *accept_context,
     {
         logWarning("file: "__FILE__", line: %d, "
                 "init_pthread_attr fail!", __LINE__);
+        return result;
     }
-    else {
-        for (i=0; i<accept_threads; i++) {
-            if ((result=pthread_create(&tid, &thread_attr,
-                            accept_thread_entrance,
-                            accept_context)) != 0)
-            {
-                logError("file: "__FILE__", line: %d, "
-                        "create thread failed, startup threads: %d, "
-                        "errno: %d, error info: %s",
-                        __LINE__, i, result, strerror(result));
-                break;
-            }
-        }
 
-        pthread_attr_destroy(&thread_attr);
+    for (i=0; i<accept_threads; i++) {
+        if ((result=pthread_create(&tid, &thread_attr,
+                        (void * (*)(void *))accept_thread_entrance,
+                        listener)) != 0)
+        {
+            logError("file: "__FILE__", line: %d, "
+                    "create thread failed, startup threads: %d, "
+                    "errno: %d, error info: %s",
+                    __LINE__, i, result, strerror(result));
+            return result;
+        }
     }
+
+    pthread_attr_destroy(&thread_attr);
+    return 0;
 }
 
-void sf_accept_loop_ex(SFContext *sf_context, const bool block)
+int sf_accept_loop_ex(SFContext *sf_context, const bool blocked)
 {
-    struct accept_thread_context *accept_contexts;
-    int count;
-    int bytes;
+    int i;
+    SFNetworkHandler *handler;
+    SFNetworkHandler *hend;
+    SFListener *listeners[SF_ADDRESS_FAMILY_COUNT *
+        SF_NETWORK_HANDLER_COUNT * 2];
+    SFListener **listener;
+    SFListener **last;
+    SFListener **lend;
 
-    if (sf_context->outer_sock >= 0) {
-        count = 2;
+    listener = listeners;
+    for (i=0; i<SF_ADDRESS_FAMILY_COUNT; i++) {
+        if (sf_context->handlers[i].af == AF_UNSPEC) {
+            continue;
+        }
+
+        hend = sf_context->handlers[i].handlers + SF_NETWORK_HANDLER_COUNT;
+        for (handler=sf_context->handlers[i].handlers;
+                handler<hend; handler++)
+        {
+            if (!handler->enabled) {
+                continue;
+            }
+
+            if (handler->inner.enabled) {
+                *listener++ = &handler->inner;
+            }
+            if (handler->outer.enabled) {
+                *listener++ = &handler->outer;
+            }
+        }
+    }
+
+    if (listener == listeners) {
+        logError("file: "__FILE__", line: %d, "
+                "no listener!", __LINE__);
+        return ENOENT;
+    }
+
+    last = listener - 1;
+    if (blocked) {
+        lend = listener - 1;
     } else {
-        count = 1;
+        lend = listener;
     }
 
-    bytes = sizeof(struct accept_thread_context) * count;
-    accept_contexts = (struct accept_thread_context *)fc_malloc(bytes);
-    if (accept_contexts == NULL) {
-        return;
+    for (listener=listeners; listener<lend; listener++) {
+        _accept_loop(*listener, sf_context->accept_threads);
     }
 
-    accept_contexts[0].sf_context = sf_context;
-    accept_contexts[0].server_sock = sf_context->inner_sock;
-
-    if (sf_context->outer_sock >= 0) {
-        accept_contexts[1].sf_context = sf_context;
-        accept_contexts[1].server_sock = sf_context->outer_sock;
-
-        if (sf_context->inner_sock >= 0) {
-            _accept_loop(accept_contexts, sf_context->accept_threads);
-        }
-
-        if (block) {
-            _accept_loop(accept_contexts + 1, sf_context->accept_threads - 1);
-            accept_thread_entrance(accept_contexts + 1);
-        } else {
-            _accept_loop(accept_contexts + 1, sf_context->accept_threads);
-        }
-    } else {
-        if (block) {
-            _accept_loop(accept_contexts, sf_context->accept_threads - 1);
-            accept_thread_entrance(accept_contexts);
-        } else {
-            _accept_loop(accept_contexts, sf_context->accept_threads);
-        }
+    if (blocked) {
+        _accept_loop(*last, sf_context->accept_threads - 1);
+        accept_run(*last);
     }
+
+    return 0;
 }
 
 #if defined(DEBUG_FLAG)
@@ -501,7 +755,7 @@ static void sigDumpHandler(int sig)
     bDumpFlag = true;
 
     snprintf(filename, sizeof(filename), 
-        "%s/logs/sf_dump.log", g_sf_global_vars.base_path);
+        "%s/logs/sf_dump.log", SF_G_BASE_PATH_STR);
     //manager_dump_global_vars_to_file(filename);
 
     bDumpFlag = false;
@@ -601,45 +855,39 @@ int sf_setup_signal_handler()
 
 int sf_startup_schedule(pthread_t *schedule_tid)
 {
-#define SCHEDULE_ENTRIES_COUNT 3
-
     ScheduleArray scheduleArray;
-    ScheduleEntry scheduleEntries[SCHEDULE_ENTRIES_COUNT];
-    int index;
+    ScheduleEntry scheduleEntries[SF_LOG_SCHEDULE_ENTRIES_COUNT];
 
     scheduleArray.entries = scheduleEntries;
-    scheduleArray.count = 0;
-
-    memset(scheduleEntries, 0, sizeof(scheduleEntries));
-
-    index = scheduleArray.count++;
-    INIT_SCHEDULE_ENTRY(scheduleEntries[index], sched_generate_next_id(),
-            TIME_NONE, TIME_NONE, 0,
-             g_sf_global_vars.sync_log_buff_interval,
-             log_sync_func, &g_log_context);
-
-    if (g_sf_global_vars.rotate_error_log) {
-        log_set_rotate_time_format(&g_log_context, "%Y%m%d");
-
-        index = scheduleArray.count++;
-        INIT_SCHEDULE_ENTRY(scheduleEntries[index], sched_generate_next_id(),
-                0, 0, 0, 86400, log_notify_rotate, &g_log_context);
-
-        if (g_sf_global_vars.log_file_keep_days > 0) {
-            log_set_keep_days(&g_log_context,
-                    g_sf_global_vars.log_file_keep_days);
-
-            index = scheduleArray.count++;
-            INIT_SCHEDULE_ENTRY(scheduleEntries[index], sched_generate_next_id(),
-                    1, 0, 0, 86400, log_delete_old_files, &g_log_context);
-        }
-    }
-
+    sf_logger_setup_schedule(&g_log_context, &g_sf_global_vars.error_log,
+            &scheduleArray);
     return sched_start(&scheduleArray, schedule_tid,
             g_sf_global_vars.thread_stack_size, (bool * volatile)
             &g_sf_global_vars.continue_flag);
 }
 
+int sf_add_slow_log_schedule(SFSlowLogContext *slowlog_ctx)
+{
+    int result;
+    ScheduleArray scheduleArray;
+    ScheduleEntry scheduleEntries[SF_LOG_SCHEDULE_ENTRIES_COUNT];
+
+    if (!slowlog_ctx->cfg.enabled) {
+        return 0;
+    }
+
+    if ((result=sf_logger_init(&slowlog_ctx->ctx, slowlog_ctx->cfg.
+                    filename_prefix)) != 0)
+    {
+        return result;
+    }
+
+    scheduleArray.entries = scheduleEntries;
+    sf_logger_setup_schedule(&slowlog_ctx->ctx, &slowlog_ctx->
+            cfg.log_cfg, &scheduleArray);
+    return sched_add_entries(&scheduleArray);
+}
+
 void sf_set_current_time()
 {
     g_current_time = time(NULL);
@@ -647,6 +895,12 @@ void sf_set_current_time()
     srand(g_sf_global_vars.up_time);
 }
 
+int sf_global_init(const char *log_filename_prefix)
+{
+    sf_set_current_time();
+    return log_set_prefix(SF_G_BASE_PATH_STR, log_filename_prefix);
+}
+
 void sf_enable_thread_notify_ex(SFContext *sf_context, const bool enabled)
 {
     struct nio_thread_data *thread_data;
@@ -668,6 +922,17 @@ struct nio_thread_data *sf_get_random_thread_data_ex(SFContext *sf_context)
     return sf_context->thread_data + index;
 }
 
+void sf_notify_all_threads_ex(SFContext *sf_context)
+{
+    struct nio_thread_data *tdata;
+    struct nio_thread_data *tend;
+
+    tend = sf_context->thread_data + sf_context->work_threads;
+    for (tdata=sf_context->thread_data; tdata<tend; tdata++) {
+        ioevent_notify_thread(tdata);
+    }
+}
+
 void sf_set_sig_quit_handler(sf_sig_quit_handler quit_handler)
 {
     sig_quit_handler = quit_handler;

src/sf_service.h

@@ -25,6 +25,9 @@
 #include "fastcommon/ioevent.h"
 #include "fastcommon/fast_task_queue.h"
 #include "sf_types.h"
+#include "sf_proto.h"
+#include "sf_global.h"
+#include "sf_nio.h"
 
 typedef void* (*sf_alloc_thread_extra_data_callback)(const int thread_index);
 typedef void (*sf_sig_quit_handler)(int sig);
@@ -33,47 +36,89 @@ typedef void (*sf_sig_quit_handler)(int sig);
 extern "C" {
 #endif
 
-int sf_service_init_ex2(SFContext *sf_context,
+int sf_service_init_ex2(SFContext *sf_context, const char *name,
         sf_alloc_thread_extra_data_callback
         alloc_thread_extra_data_callback,
         ThreadLoopCallback thread_loop_callback,
         sf_accept_done_callback accept_done_callback,
         sf_set_body_length_callback set_body_length_func,
-        sf_deal_task_func deal_func, TaskCleanUpCallback task_cleanup_func,
+        sf_alloc_recv_buffer_callback alloc_recv_buffer_func,
+        sf_send_done_callback send_done_callback,
+        sf_deal_task_callback deal_func, TaskCleanUpCallback task_cleanup_func,
         sf_recv_timeout_callback timeout_callback, const int net_timeout_ms,
-        const int proto_header_size, const int task_arg_size,
-        TaskInitCallback init_callback);
+        const int proto_header_size, const int task_padding_size,
+        const int task_arg_size, const bool double_buffers,
+        const bool need_shrink_task_buffer, const bool explicit_post_recv,
+        TaskInitCallback init_callback, void *init_arg,
+        sf_release_buffer_callback release_buffer_callback);
 
-#define sf_service_init_ex(sf_context, alloc_thread_extra_data_callback,  \
-        thread_loop_callback, accept_done_callback, set_body_length_func, \
-        deal_func, task_cleanup_func, timeout_callback, net_timeout_ms,   \
-        proto_header_size, task_arg_size) \
-    sf_service_init_ex2(sf_context, alloc_thread_extra_data_callback,     \
-        thread_loop_callback, accept_done_callback, set_body_length_func, \
-        deal_func, task_cleanup_func, timeout_callback, net_timeout_ms,   \
-        proto_header_size, task_arg_size, NULL)
+#define sf_service_init_ex(sf_context, name, alloc_thread_extra_data_callback,\
+        thread_loop_callback, accept_done_callback, set_body_length_func,   \
+        send_done_callback, deal_func, task_cleanup_func, timeout_callback, \
+        net_timeout_ms, proto_header_size, task_arg_size) \
+    sf_service_init_ex2(sf_context, name, alloc_thread_extra_data_callback,   \
+        thread_loop_callback, accept_done_callback, set_body_length_func,     \
+        NULL, send_done_callback, deal_func, task_cleanup_func, \
+        timeout_callback, net_timeout_ms, proto_header_size, \
+        0, task_arg_size, false, true, false, NULL, NULL, NULL)
 
-#define sf_service_init(alloc_thread_extra_data_callback, \
-        thread_loop_callback, accept_done_callback, set_body_length_func, \
-        deal_func, task_cleanup_func, timeout_callback, net_timeout_ms,   \
-        proto_header_size, task_arg_size) \
-    sf_service_init_ex2(&g_sf_context, alloc_thread_extra_data_callback,  \
-        thread_loop_callback, accept_done_callback, set_body_length_func, \
-        deal_func, task_cleanup_func, timeout_callback, net_timeout_ms,   \
-        proto_header_size, task_arg_size, NULL)
+#define sf_service_init(name, alloc_thread_extra_data_callback, \
+        thread_loop_callback, accept_done_callback, set_body_length_func,   \
+        send_done_callback, deal_func, task_cleanup_func, timeout_callback, \
+        net_timeout_ms, proto_header_size, task_arg_size) \
+    sf_service_init_ex2(&g_sf_context, name, alloc_thread_extra_data_callback, \
+        thread_loop_callback, accept_done_callback, set_body_length_func, NULL,\
+        send_done_callback, deal_func, task_cleanup_func, timeout_callback, \
+        net_timeout_ms, proto_header_size, 0, task_arg_size, false, true,   \
+        false, NULL, NULL, NULL)
 
 int sf_service_destroy_ex(SFContext *sf_context);
 
 #define sf_service_destroy() sf_service_destroy_ex(&g_sf_context)
 
+void sf_service_set_thread_loop_callback_ex(SFContext *sf_context,
+        ThreadLoopCallback thread_loop_callback);
+
+#define sf_service_set_thread_loop_callback(thread_loop_callback) \
+    sf_service_set_thread_loop_callback_ex(&g_sf_context, thread_loop_callback)
+
+static inline void sf_service_set_smart_polling_ex(SFContext *sf_context,
+        const FCSmartPollingConfig *smart_polling)
+{
+    sf_context->smart_polling = *smart_polling;
+}
+#define sf_service_set_smart_polling(smart_polling) \
+    sf_service_set_smart_polling_ex(&g_sf_context, smart_polling)
+
+static inline void sf_service_set_connect_need_log_ex(
+        SFContext *sf_context, const bool need_log)
+{
+    sf_context->connect_need_log = need_log;
+}
+#define sf_service_set_connect_need_log(need_log) \
+    sf_service_set_connect_need_log_ex(&g_sf_context, need_log)
+
+
 int sf_setup_signal_handler();
+
 int sf_startup_schedule(pthread_t *schedule_tid);
+int sf_add_slow_log_schedule(SFSlowLogContext *slowlog_ctx);
+
 void sf_set_current_time();
+int sf_global_init(const char *log_filename_prefix);
+
+int sf_socket_create_server(SFListener *listener,
+        int af, const char *bind_addr);
+void sf_socket_close_server(SFListener *listener);
+struct fast_task_info *sf_socket_accept_connection(SFListener *listener);
 
 int sf_socket_server_ex(SFContext *sf_context);
 #define sf_socket_server() sf_socket_server_ex(&g_sf_context)
 
-void sf_accept_loop_ex(SFContext *sf_context, const bool block);
+void sf_socket_close_ex(SFContext *sf_context);
+#define sf_socket_close() sf_socket_close_ex(&g_sf_context)
+
+int sf_accept_loop_ex(SFContext *sf_context, const bool blocked);
 
 #define sf_accept_loop()  sf_accept_loop_ex(&g_sf_context, true)
 
@@ -96,16 +141,22 @@ struct nio_thread_data *sf_get_random_thread_data_ex(SFContext *sf_context);
 #define sf_get_random_thread_data()  \
     sf_get_random_thread_data_ex(&g_sf_context)
 
+
+void sf_notify_all_threads_ex(SFContext *sf_context);
+
+#define sf_notify_all_threads()  \
+    sf_notify_all_threads_ex(&g_sf_context)
+
+
 void sf_set_sig_quit_handler(sf_sig_quit_handler quit_handler);
 
-int sf_init_task(struct fast_task_info *task);
-
-static inline struct fast_task_info *sf_alloc_init_task(
-        SFContext *sf_context, const int sock)
+static inline struct fast_task_info *sf_alloc_init_task_ex(
+        SFNetworkHandler *handler, const int fd,
+        const int reffer_count)
 {
     struct fast_task_info *task;
 
-    task = free_queue_pop();
+    task = free_queue_pop(&handler->fh->ctx->free_queue);
     if (task == NULL) {
         logError("file: "__FILE__", line: %d, "
                 "malloc task buff failed, you should "
@@ -113,19 +164,57 @@ static inline struct fast_task_info *sf_alloc_init_task(
                 __LINE__);
         return NULL;
     }
-    __sync_add_and_fetch(&task->reffer_count, 1);
+
+    if (task->shrinked) {
+        task->shrinked = false;
+        sf_proto_init_task_magic(task);
+    }
+
+    __sync_add_and_fetch(&task->reffer_count, reffer_count);
     __sync_bool_compare_and_swap(&task->canceled, 1, 0);
-    task->ctx = sf_context;
-    task->event.fd = sock;
+    task->handler = handler;
+    task->event.fd = fd;
+    return task;
+}
+
+#define sf_hold_task_ex(task, inc_count)  fc_hold_task_ex(task, inc_count)
+#define sf_hold_task(task)  fc_hold_task(task)
+
+#define sf_alloc_init_task(handler, fd) sf_alloc_init_task_ex(handler, fd, 1)
+
+static inline struct fast_task_info *sf_alloc_init_server_task(
+        SFNetworkHandler *handler, const int fd)
+{
+    const int reffer_count = 1;
+    struct fast_task_info *task;
+
+    if ((task=sf_alloc_init_task_ex(handler, fd, reffer_count)) != NULL) {
+#if IOEVENT_USE_URING
+        FC_URING_IS_CLIENT(task) = false;
+#endif
+    }
 
     return task;
 }
 
-#define sf_hold_task(task) __sync_add_and_fetch(&task->reffer_count, 1)
+static inline struct fast_task_info *sf_alloc_init_client_task(
+        SFNetworkHandler *handler)
+{
+    const int fd = -1;
+    const int reffer_count = 1;
+    struct fast_task_info *task;
+
+    if ((task=sf_alloc_init_task_ex(handler, fd, reffer_count)) != NULL) {
+#if IOEVENT_USE_URING
+        FC_URING_IS_CLIENT(task) = true;
+#endif
+    }
+
+    return task;
+}
 
 static inline void sf_release_task(struct fast_task_info *task)
 {
-    //int reffer_count;
     if (__sync_sub_and_fetch(&task->reffer_count, 1) == 0) {
         /*
         int free_count = free_queue_count();
@@ -136,13 +225,99 @@ static inline void sf_release_task(struct fast_task_info *task)
                 alloc_count, alloc_count - free_count, free_count);
                 */
 
+#if IOEVENT_USE_URING
+        if (SF_CTX->use_io_uring) {
+            task->handler->close_connection(task);
+            __sync_fetch_and_sub(&g_sf_global_vars.
+                    connection_stat.current_count, 1);
+        }
+#endif
+
         free_queue_push(task);
-    } else {
-        /*
-        logInfo("file: "__FILE__", line: %d, "
-                "release task %p, current reffer: %d",
-                __LINE__, task, reffer_count);
-                */
+    }
+}
+
+static inline SFNetworkHandler *sf_get_first_network_handler_ex(
+        SFContext *sf_context)
+{
+    int i;
+    SFNetworkHandler *handler;
+    SFNetworkHandler *end;
+
+    for (i=0; i<SF_ADDRESS_FAMILY_COUNT; i++) {
+        if (sf_context->handlers[i].af == AF_UNSPEC) {
+            continue;
+        }
+
+        end = sf_context->handlers[i].handlers + SF_NETWORK_HANDLER_COUNT;
+        for (handler=sf_context->handlers[i].handlers; handler<end; handler++) {
+            if (handler->enabled) {
+                return handler;
+            }
+        }
+    }
+
+    return NULL;
+}
+#define sf_get_first_network_handler() \
+    sf_get_first_network_handler_ex(&g_sf_context)
+
+
+static inline SFNetworkHandler *sf_get_rdma_network_handler(
+        SFContext *sf_context)
+{
+    int i;
+    SFNetworkHandler *handler;
+
+    for (i=0; i<SF_ADDRESS_FAMILY_COUNT; i++) {
+        if (sf_context->handlers[i].af != AF_UNSPEC) {
+            handler = sf_context->handlers[i].handlers +
+                SF_RDMACM_NETWORK_HANDLER_INDEX;
+            if (handler->enabled) {
+                return handler;
+            }
+        }
+    }
+
+    return NULL;
+}
+
+static inline SFNetworkHandler *sf_get_rdma_network_handler2(
+        SFContext *sf_context1, SFContext *sf_context2)
+{
+    SFNetworkHandler *handler;
+
+    if ((handler=sf_get_rdma_network_handler(sf_context1)) != NULL) {
+        return handler;
+    }
+    return sf_get_rdma_network_handler(sf_context2);
+}
+
+static inline SFNetworkHandler *sf_get_rdma_network_handler3(
+        SFContext *sf_context1, SFContext *sf_context2,
+        SFContext *sf_context3)
+{
+    SFNetworkHandler *handler;
+
+    if ((handler=sf_get_rdma_network_handler(sf_context1)) != NULL) {
+        return handler;
+    }
+    if ((handler=sf_get_rdma_network_handler(sf_context2)) != NULL) {
+        return handler;
+    }
+    return sf_get_rdma_network_handler(sf_context3);
+}
+
+static inline bool sf_get_double_buffers_flag(FCServerGroupInfo *server_group)
+{
+    if (server_group->comm_type == fc_comm_type_sock) {
+#if IOEVENT_USE_URING
+        return true;
+#else
+        return false;
+#endif
+    } else {  //RDMA
+        return true;
     }
 }
 

src/sf_sharding_htable.c

@@ -15,6 +15,7 @@
 
 #include <stdlib.h>
 #include "fastcommon/shared_func.h"
+#include "fastcommon/fc_atomic.h"
 #include "sf_sharding_htable.h"
 
 static int init_allocators(SFHtableShardingContext *sharding_ctx,
@@ -44,7 +45,7 @@ static int init_allocators(SFHtableShardingContext *sharding_ctx,
 
     end = sharding_ctx->allocators.elts + allocator_count;
     for (pa=sharding_ctx->allocators.elts; pa<end; pa++) {
-        if ((result=fast_mblock_init_ex1(pa, "sharding_hkey", element_size,
+        if ((result=fast_mblock_init_ex1(pa, "sharding-hkey", element_size,
                         alloc_elts_once, 0, NULL, NULL, true)) != 0)
         {
             return result;
@@ -78,7 +79,7 @@ static int init_sharding(SFHtableSharding *sharding,
 
     sharding->hashtable.capacity = per_capacity;
     sharding->element_count = 0;
-    sharding->last_reclaim_time_sec = get_current_time();
+    sharding->last_reclaim_time_ms = 1000LL * (int64_t)get_current_time();
     FC_INIT_LIST_HEAD(&sharding->lru);
     return 0;
 }
@@ -114,15 +115,16 @@ static int init_sharding_array(SFHtableShardingContext *sharding_ctx,
     return 0;
 }
 
-int sf_sharding_htable_init(SFHtableShardingContext *sharding_ctx,
+int sf_sharding_htable_init_ex(SFHtableShardingContext *sharding_ctx,
         const SFShardingHtableKeyType key_type,
         sf_sharding_htable_insert_callback insert_callback,
         sf_sharding_htable_find_callback find_callback,
+        sf_sharding_htable_delete_callback delete_callback,
         sf_sharding_htable_accept_reclaim_callback reclaim_callback,
         const int sharding_count, const int64_t htable_capacity,
         const int allocator_count, const int element_size,
-        int64_t element_limit, const int64_t min_ttl_sec,
-        const int64_t max_ttl_sec)
+        int64_t element_limit, const int64_t min_ttl_ms,
+        const int64_t max_ttl_ms, const double low_water_mark_ratio)
 {
     int result;
     int64_t per_elt_limit;
@@ -131,6 +133,7 @@ int sf_sharding_htable_init(SFHtableShardingContext *sharding_ctx,
     if (element_limit <= 0) {
         element_limit = 1000 * 1000;
     }
+
     if ((result=init_allocators(sharding_ctx, allocator_count,
                     element_size, element_limit)) != 0)
     {
@@ -148,19 +151,23 @@ int sf_sharding_htable_init(SFHtableShardingContext *sharding_ctx,
     sharding_ctx->key_type = key_type;
     sharding_ctx->insert_callback = insert_callback;
     sharding_ctx->find_callback = find_callback;
+    sharding_ctx->delete_callback = delete_callback;
     sharding_ctx->accept_reclaim_callback = reclaim_callback;
-    sharding_ctx->sharding_reclaim.elt_water_mark = per_elt_limit * 0.10;
-    sharding_ctx->sharding_reclaim.min_ttl_sec = min_ttl_sec;
-    sharding_ctx->sharding_reclaim.max_ttl_sec = max_ttl_sec;
-    sharding_ctx->sharding_reclaim.elt_ttl_sec = (double)(sharding_ctx->
-            sharding_reclaim.max_ttl_sec - sharding_ctx->
-            sharding_reclaim.min_ttl_sec) / per_elt_limit;
+
+    sharding_ctx->sharding_reclaim.enabled = (delete_callback == NULL);
+    sharding_ctx->sharding_reclaim.elt_water_mark =
+        per_elt_limit * low_water_mark_ratio;
+    sharding_ctx->sharding_reclaim.min_ttl_ms = min_ttl_ms;
+    sharding_ctx->sharding_reclaim.max_ttl_ms = max_ttl_ms;
+    sharding_ctx->sharding_reclaim.elt_ttl_ms = (double)(sharding_ctx->
+            sharding_reclaim.max_ttl_ms - sharding_ctx->
+            sharding_reclaim.min_ttl_ms) / per_elt_limit;
 
     /*
     logInfo("per_elt_limit: %"PRId64", elt_water_mark: %d, "
-            "elt_ttl_sec: %.2f", per_elt_limit, (int)sharding_ctx->
+            "elt_ttl_ms: %.2f", per_elt_limit, (int)sharding_ctx->
             sharding_reclaim.elt_water_mark, sharding_ctx->
-            sharding_reclaim.elt_ttl_sec);
+            sharding_reclaim.elt_ttl_ms);
             */
     return 0;
 }
@@ -220,9 +227,10 @@ static inline void htable_insert(SFHtableShardingContext *sharding_ctx,
     fc_list_add_internal(&entry->dlinks.htable, previous, previous->next);
 }
 
-static SFShardingHashEntry *otid_entry_reclaim(SFHtableSharding *sharding)
+static SFShardingHashEntry *hash_entry_reclaim(SFHtableSharding *sharding)
 {
-    int64_t reclaim_ttl_sec;
+    int64_t current_time_ms;
+    int64_t reclaim_ttl_ms;
     int64_t delta;
     int64_t reclaim_count;
     int64_t reclaim_limit;
@@ -232,7 +240,14 @@ static SFShardingHashEntry *otid_entry_reclaim(SFHtableSharding *sharding)
 
     if (sharding->element_count <= sharding->element_limit) {
         delta = sharding->element_count;
-        reclaim_limit = sharding->ctx->sharding_reclaim.elt_water_mark;
+        if (sharding->ctx->sharding_reclaim.elt_water_mark > 0) {
+            reclaim_count = sharding->element_count - sharding->ctx->
+                sharding_reclaim.elt_water_mark;
+            reclaim_limit = FC_MIN(reclaim_count, sharding->ctx->
+                    sharding_reclaim.elt_water_mark);
+        } else {
+            reclaim_limit = sharding->element_count;
+        }
     } else {
         delta = sharding->element_limit;
         reclaim_limit = (sharding->element_count - sharding->element_limit) +
@@ -241,12 +256,11 @@ static SFShardingHashEntry *otid_entry_reclaim(SFHtableSharding *sharding)
 
     first = NULL;
     reclaim_count = 0;
-    reclaim_ttl_sec = (int64_t)(sharding->ctx->sharding_reclaim.max_ttl_sec -
-            sharding->ctx->sharding_reclaim.elt_ttl_sec * delta);
+    current_time_ms = 1000LL * (int64_t)get_current_time();
+    reclaim_ttl_ms = (int64_t)(sharding->ctx->sharding_reclaim.max_ttl_ms -
+            sharding->ctx->sharding_reclaim.elt_ttl_ms * delta);
     fc_list_for_each_entry_safe(entry, tmp, &sharding->lru, dlinks.lru) {
-        if (get_current_time() - entry->last_update_time_sec <=
-                reclaim_ttl_sec)
-        {
+        if (current_time_ms - entry->last_update_time_ms <= reclaim_ttl_ms) {
             break;
         }
 
@@ -271,28 +285,37 @@ static SFShardingHashEntry *otid_entry_reclaim(SFHtableSharding *sharding)
     }
 
     if (reclaim_count > 0) {
-        logInfo("sharding index: %d, element_count: %"PRId64", "
-                "reclaim_ttl_sec: %"PRId64" ms, reclaim_count: %"PRId64", "
+        logDebug("sharding index: %d, element_count: %"PRId64", "
+                "reclaim_ttl_ms: %"PRId64" ms, reclaim_count: %"PRId64", "
                 "reclaim_limit: %"PRId64, (int)(sharding - sharding->ctx->
                     sharding_array.entries), sharding->element_count,
-                reclaim_ttl_sec, reclaim_count, reclaim_limit);
+                reclaim_ttl_ms, reclaim_count, reclaim_limit);
     }
 
     return first;
 }
 
 static inline SFShardingHashEntry *htable_entry_alloc(
+        SFHtableShardingContext *sharding_ctx,
         SFHtableSharding *sharding)
 {
     SFShardingHashEntry *entry;
+    int64_t current_time_ms;
+    int64_t last_reclaim_time_ms;
 
-    if (sharding->element_count > sharding->ctx->sharding_reclaim.
-            elt_water_mark && get_current_time() - sharding->
-            last_reclaim_time_sec > 1000)
+    if (sharding_ctx->sharding_reclaim.enabled &&
+            (sharding->element_count > sharding->ctx->
+             sharding_reclaim.elt_water_mark))
     {
-        sharding->last_reclaim_time_sec = get_current_time();
-        if ((entry=otid_entry_reclaim(sharding)) != NULL) {
-            return entry;
+        current_time_ms = 1000LL * (int64_t)get_current_time();
+        last_reclaim_time_ms = FC_ATOMIC_GET(sharding->last_reclaim_time_ms);
+        if (current_time_ms - last_reclaim_time_ms > 100 &&
+                __sync_bool_compare_and_swap(&sharding->last_reclaim_time_ms,
+                    last_reclaim_time_ms, current_time_ms))
+        {
+            if ((entry=hash_entry_reclaim(sharding)) != NULL) {
+                return entry;
+            }
         }
     }
 
@@ -309,7 +332,6 @@ static inline SFShardingHashEntry *htable_entry_alloc(
 #define SET_SHARDING_AND_BUCKET(sharding_ctx, key) \
     SFHtableSharding *sharding; \
     struct fc_list_head *bucket;   \
-    SFShardingHashEntry *entry; \
     uint64_t hash_code;    \
     \
     hash_code = sf_sharding_htable_key_ids_one == sharding_ctx-> \
@@ -324,6 +346,7 @@ void *sf_sharding_htable_find(SFHtableShardingContext
         *sharding_ctx, const SFTwoIdsHashKey *key, void *arg)
 {
     void *data;
+    SFShardingHashEntry *entry;
     SET_SHARDING_AND_BUCKET(sharding_ctx, key);
 
     PTHREAD_MUTEX_LOCK(&sharding->lock);
@@ -338,9 +361,43 @@ void *sf_sharding_htable_find(SFHtableShardingContext
     return data;
 }
 
+int sf_sharding_htable_delete(SFHtableShardingContext
+        *sharding_ctx, const SFTwoIdsHashKey *key, void *arg)
+{
+    int result;
+    SFShardingHashEntry *entry;
+
+    if (sharding_ctx->delete_callback != NULL) {
+        SET_SHARDING_AND_BUCKET(sharding_ctx, key);
+        PTHREAD_MUTEX_LOCK(&sharding->lock);
+        entry = htable_find(sharding_ctx, key, bucket);
+        if (entry != NULL) {
+            if (sharding_ctx->delete_callback(entry, arg)) {
+                fc_list_del_init(&entry->dlinks.htable);
+                if (sharding_ctx->sharding_reclaim.enabled) {
+                    fc_list_del_init(&entry->dlinks.lru);
+                }
+                fast_mblock_free_object(sharding->allocator, entry);
+                sharding->element_count--;
+            }
+            result = 0;
+        } else {
+            result = ENOENT;
+        }
+        PTHREAD_MUTEX_UNLOCK(&sharding->lock);
+    } else {
+        logError("file: "__FILE__", line: %d, "
+                "delete callback is NULL!", __LINE__);
+        result = EINVAL;
+    }
+
+    return result;
+}
+
 int sf_sharding_htable_insert(SFHtableShardingContext
         *sharding_ctx, const SFTwoIdsHashKey *key, void *arg)
 {
+    SFShardingHashEntry *entry;
     bool new_create;
     int result;
     SET_SHARDING_AND_BUCKET(sharding_ctx, key);
@@ -348,7 +405,7 @@ int sf_sharding_htable_insert(SFHtableShardingContext
     PTHREAD_MUTEX_LOCK(&sharding->lock);
     do {
         if ((entry=htable_find(sharding_ctx, key, bucket)) == NULL) {
-            if ((entry=htable_entry_alloc(sharding)) == NULL) {
+            if ((entry=htable_entry_alloc(sharding_ctx, sharding)) == NULL) {
                 result = ENOMEM;
                 break;
             }
@@ -356,13 +413,17 @@ int sf_sharding_htable_insert(SFHtableShardingContext
             new_create = true;
             entry->key = *key;
             htable_insert(sharding_ctx, entry, bucket);
-            fc_list_add_tail(&entry->dlinks.lru, &sharding->lru);
+            if (sharding_ctx->sharding_reclaim.enabled) {
+                fc_list_add_tail(&entry->dlinks.lru, &sharding->lru);
+            }
         } else {
             new_create = false;
-            fc_list_move_tail(&entry->dlinks.lru, &sharding->lru);
+            if (sharding_ctx->sharding_reclaim.enabled) {
+                fc_list_move_tail(&entry->dlinks.lru, &sharding->lru);
+            }
         }
 
-        entry->last_update_time_sec = get_current_time();
+        entry->last_update_time_ms = 1000LL * (int64_t)get_current_time();
         result = sharding_ctx->insert_callback(
                 entry, arg, new_create);
     } while (0);

src/sf_sharding_htable.h

@@ -18,7 +18,6 @@
 
 #include <limits.h>
 #include <sys/types.h>
-#include <sys/stat.h>
 #include "fastcommon/common_define.h"
 #include "fastcommon/fc_list.h"
 #include "fastcommon/pthread_func.h"
@@ -37,6 +36,9 @@ typedef int (*sf_sharding_htable_insert_callback)
 typedef void *(*sf_sharding_htable_find_callback)
     (struct sf_sharding_hash_entry *entry, void *arg);
 
+typedef bool (*sf_sharding_htable_delete_callback)
+    (struct sf_sharding_hash_entry *entry, void *arg);
+
 typedef bool (*sf_sharding_htable_accept_reclaim_callback)
     (struct sf_sharding_hash_entry *entry);
 
@@ -59,7 +61,7 @@ typedef struct sf_sharding_hash_entry {
         struct fc_list_head htable;  //for hashtable
         struct fc_list_head lru;     //for LRU chain
     } dlinks;
-    int64_t last_update_time_sec;
+    int64_t last_update_time_ms;
     struct sf_htable_sharding *sharding;  //hold for lock
 } SFShardingHashEntry;
 
@@ -76,7 +78,7 @@ typedef struct sf_htable_sharding {
     SFDlinkHashtable hashtable;
     int64_t element_count;
     int64_t element_limit;
-    int64_t last_reclaim_time_sec;
+    volatile int64_t last_reclaim_time_ms;
     struct sf_htable_sharding_context *ctx;
 } SFHtableSharding;
 
@@ -87,20 +89,22 @@ typedef struct sf_htable_sharding_array {
 
 typedef struct sf_htable_sharding_context {
     struct {
-        int64_t min_ttl_sec;
-        int64_t max_ttl_sec;
-        double elt_ttl_sec;
+        int64_t min_ttl_ms;
+        int64_t max_ttl_ms;
+        double elt_ttl_ms;
         int elt_water_mark;  //trigger reclaim when elements exceeds water mark
+        bool enabled;
     } sharding_reclaim;
 
     struct {
         int count;
         struct fast_mblock_man *elts;
-    } allocators;
+    } allocators;  //shared allocators
 
     SFShardingHtableKeyType key_type;  //id count in the hash entry
     sf_sharding_htable_insert_callback insert_callback;
     sf_sharding_htable_find_callback find_callback;
+    sf_sharding_htable_delete_callback delete_callback;
     sf_sharding_htable_accept_reclaim_callback accept_reclaim_callback;
     SFHtableShardingArray sharding_array;
 } SFHtableShardingContext;
@@ -109,15 +113,35 @@ typedef struct sf_htable_sharding_context {
 extern "C" {
 #endif
 
-    int sf_sharding_htable_init(SFHtableShardingContext *sharding_ctx,
+    int sf_sharding_htable_init_ex(SFHtableShardingContext *sharding_ctx,
             const SFShardingHtableKeyType key_type,
             sf_sharding_htable_insert_callback insert_callback,
             sf_sharding_htable_find_callback find_callback,
+            sf_sharding_htable_delete_callback delete_callback,
             sf_sharding_htable_accept_reclaim_callback reclaim_callback,
             const int sharding_count, const int64_t htable_capacity,
             const int allocator_count, const int element_size,
-            int64_t element_limit, const int64_t min_ttl_sec,
-            const int64_t max_ttl_sec);
+            int64_t element_limit, const int64_t min_ttl_ms,
+            const int64_t max_ttl_ms, const double low_water_mark_ratio);
+
+    static inline int sf_sharding_htable_init(SFHtableShardingContext
+            *sharding_ctx, const SFShardingHtableKeyType key_type,
+            sf_sharding_htable_insert_callback insert_callback,
+            sf_sharding_htable_find_callback find_callback,
+            sf_sharding_htable_delete_callback delete_callback,
+            sf_sharding_htable_accept_reclaim_callback reclaim_callback,
+            const int sharding_count, const int64_t htable_capacity,
+            const int allocator_count, const int element_size,
+            int64_t element_limit, const int64_t min_ttl_ms,
+            const int64_t max_ttl_ms)
+    {
+        const double low_water_mark_ratio = 0.10;
+        return sf_sharding_htable_init_ex(sharding_ctx, key_type,
+                insert_callback, find_callback, delete_callback,
+                reclaim_callback, sharding_count, htable_capacity,
+                allocator_count, element_size, element_limit,
+                min_ttl_ms, max_ttl_ms, low_water_mark_ratio);
+    }
 
     int sf_sharding_htable_insert(SFHtableShardingContext
             *sharding_ctx, const SFTwoIdsHashKey *key, void *arg);
@@ -125,6 +149,9 @@ extern "C" {
     void *sf_sharding_htable_find(SFHtableShardingContext
             *sharding_ctx, const SFTwoIdsHashKey *key, void *arg);
 
+    int sf_sharding_htable_delete(SFHtableShardingContext
+            *sharding_ctx, const SFTwoIdsHashKey *key, void *arg);
+
 #ifdef __cplusplus
 }
 #endif

src/sf_shared_mbuffer.c

@@ -0,0 +1,94 @@
+/*
+ * Copyright (c) 2020 YuQing <384681@qq.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the Lesser GNU General Public License, version 3
+ * or later ("LGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the Lesser GNU General Public License
+ * along with this program. If not, see <https://www.gnu.org/licenses/>.
+ */
+
+#include "sf_shared_mbuffer.h"
+
+static int sf_shared_mbuffer_alloc_init(void *element, void *args)
+{
+    SFSharedMBuffer *buffer;
+
+    buffer = (SFSharedMBuffer *)((char *)element +
+            sizeof(struct fast_allocator_wrapper));
+    buffer->ctx = (SFSharedMBufferContext *)args;
+    return 0;
+}
+
+int sf_shared_mbuffer_init_ex(SFSharedMBufferContext *context,
+        const char *name_prefix, const int buff_extra_size,
+        const int min_buff_size, const int max_buff_size,
+        const int min_alloc_once, const int64_t memory_limit,
+        const bool need_lock)
+{
+    const double expect_usage_ratio = 0.75;
+    const int reclaim_interval = 1;
+    struct fast_region_info regions[32];
+    struct fast_mblock_object_callbacks object_callbacks;
+    int count;
+    int start;
+    int end;
+    int alloc_once;
+    int buff_size;
+    int i;
+
+    alloc_once = (4 * 1024 * 1024) / max_buff_size;
+    if (alloc_once == 0) {
+        alloc_once = min_alloc_once;
+    } else {
+        i = min_alloc_once;
+        while (i < alloc_once) {
+            i *= 2;
+        }
+        alloc_once = i;
+    }
+
+    count = 1;
+    buff_size = min_buff_size;
+    while (buff_size < max_buff_size) {
+        buff_size *= 2;
+        ++count;
+        alloc_once *= 2;
+    }
+
+    buff_size = min_buff_size;
+    start = 0;
+    end = buff_extra_size + buff_size;
+    FAST_ALLOCATOR_INIT_REGION(regions[0], start, end,
+            end - start, alloc_once);
+
+    //logInfo("[1] start: %d, end: %d, alloc_once: %d", start, end, alloc_once);
+
+    start = end;
+    for (i=1; i<count; i++) {
+        buff_size *= 2;
+        alloc_once /= 2;
+        end = buff_extra_size + buff_size;
+        FAST_ALLOCATOR_INIT_REGION(regions[i], start, end,
+                end - start, alloc_once);
+        //logInfo("[%d] start: %d, end: %d, alloc_once: %d", i + 1, start, end, alloc_once);
+        start = end;
+    }
+
+    object_callbacks.init_func = sf_shared_mbuffer_alloc_init;
+    object_callbacks.destroy_func = NULL;
+    object_callbacks.args = context;
+    return fast_allocator_init_ex(&context->allocator, name_prefix,
+            sizeof(SFSharedMBuffer), &object_callbacks, regions, count,
+            memory_limit, expect_usage_ratio, reclaim_interval, need_lock);
+}
+
+void sf_shared_mbuffer_destroy(SFSharedMBufferContext *context)
+{
+    fast_allocator_destroy(&context->allocator);
+}

src/sf_shared_mbuffer.h

@@ -0,0 +1,114 @@
+/*
+ * Copyright (c) 2020 YuQing <384681@qq.com>
+ *
+ * This program is free software: you can use, redistribute, and/or modify
+ * it under the terms of the Lesser GNU General Public License, version 3
+ * or later ("LGPL"), as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * You should have received a copy of the Lesser GNU General Public License
+ * along with this program. If not, see <https://www.gnu.org/licenses/>.
+ */
+
+#ifndef _SF_SHARED_MBUFFER_H__
+#define _SF_SHARED_MBUFFER_H__
+
+#include "fastcommon/fc_list.h"
+#include "fastcommon/fast_task_queue.h"
+#include "fastcommon/shared_func.h"
+#include "fastcommon/logger.h"
+#include "fastcommon/fast_allocator.h"
+
+typedef struct sf_shared_mbuffer_context {
+    struct fast_allocator_context allocator;
+} SFSharedMBufferContext;
+
+typedef struct sf_shared_mbuffer {
+    int length;
+    volatile int reffer_count;
+    SFSharedMBufferContext *ctx;
+    char buff[0];  //must be last
+} SFSharedMBuffer;
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#define sf_shared_mbuffer_init(context, name_prefix, buff_extra_size, \
+        min_buff_size, max_buff_size, min_alloc_once, memory_limit)   \
+    sf_shared_mbuffer_init_ex(context, name_prefix, buff_extra_size,  \
+        min_buff_size, max_buff_size, min_alloc_once, memory_limit, true)
+
+int sf_shared_mbuffer_init_ex(SFSharedMBufferContext *context,
+        const char *name_prefix, const int buff_extra_size,
+        const int min_buff_size, const int max_buff_size,
+        const int min_alloc_once, const int64_t memory_limit,
+        const bool need_lock);
+
+void sf_shared_mbuffer_destroy(SFSharedMBufferContext *context);
+
+#define sf_shared_mbuffer_alloc(context, buffer_size)  \
+    sf_shared_mbuffer_alloc_ex(context, buffer_size, 1)
+
+static inline SFSharedMBuffer *sf_shared_mbuffer_alloc_ex(
+        SFSharedMBufferContext *context, const int buffer_size,
+        const int init_reffer_count)
+{
+    SFSharedMBuffer *buffer;
+    int sleep_ms;
+
+    sleep_ms = 5;
+    while ((buffer=fast_allocator_alloc(&context->allocator,
+                    buffer_size)) == NULL)
+    {
+        if (sleep_ms < 100) {
+            sleep_ms *= 2;
+        }
+        fc_sleep_ms(sleep_ms);
+    }
+
+    if (init_reffer_count > 0) {
+        __sync_add_and_fetch(&buffer->reffer_count, init_reffer_count);
+    }
+
+    /*
+    logInfo("file: "__FILE__", line: %d, "
+            "alloc shared buffer: %p, buff: %p, reffer_count: %d",
+            __LINE__, buffer, buffer->buff, __sync_add_and_fetch(&buffer->reffer_count, 0));
+            */
+
+    return buffer;
+}
+
+static inline void sf_shared_mbuffer_hold(SFSharedMBuffer *buffer)
+{
+    __sync_add_and_fetch(&buffer->reffer_count, 1);
+}
+
+static inline void sf_shared_mbuffer_release(SFSharedMBuffer *buffer)
+{
+    if (__sync_sub_and_fetch(&buffer->reffer_count, 1) == 0) {
+        /*
+        logInfo("file: "__FILE__", line: %d, "
+                "free shared buffer: %p", __LINE__, buffer);
+                */
+        fast_allocator_free(&buffer->ctx->allocator, buffer);
+    }
+}
+
+static inline void sf_release_task_shared_mbuffer(struct fast_task_info *task)
+{
+    SFSharedMBuffer *mbuffer;
+    mbuffer = fc_list_entry(task->recv_body, SFSharedMBuffer, buff);
+    sf_shared_mbuffer_release(mbuffer);
+    task->recv_body = NULL;
+}
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif

src/sf_types.h

@@ -25,47 +25,185 @@
 #include <time.h>
 #include "fastcommon/connection_pool.h"
 #include "fastcommon/fast_task_queue.h"
+#include "fastcommon/server_id_func.h"
 
 #define SF_ERROR_INFO_SIZE   256
+#define SF_CLUSTER_CONFIG_SIGN_LEN  16
 
 #define SF_SERVER_TASK_TYPE_NONE                 0
 #define SF_SERVER_TASK_TYPE_CHANNEL_HOLDER     101   //for request idempotency
 #define SF_SERVER_TASK_TYPE_CHANNEL_USER       102   //for request idempotency
 
-typedef void (*sf_accept_done_callback)(struct fast_task_info *task,
-        const bool bInnerPort);
+#define SF_ADDRESS_FAMILY_COUNT           2
+#define SF_IPV4_ADDRESS_FAMILY_INDEX      0
+#define SF_IPV6_ADDRESS_FAMILY_INDEX      1
+
+#define SF_NETWORK_HANDLER_COUNT          2
+#define SF_SOCKET_NETWORK_HANDLER_INDEX   0
+#define SF_RDMACM_NETWORK_HANDLER_INDEX   1
+
+#define SF_BINLOG_BUFFER_PRODUCER_DATA_LENGTH(bf) ((bf).data_end - (bf).buff)
+#define SF_BINLOG_BUFFER_PRODUCER_BUFF_REMAIN(bf) ((bf).buff_end - (bf).data_end)
+#define SF_BINLOG_BUFFER_CONSUMER_DATA_LENGTH(bf) ((bf).current  - (bf).buff)
+#define SF_BINLOG_BUFFER_CONSUMER_DATA_REMAIN(bf) ((bf).data_end - (bf).current)
+
+typedef int (*sf_accept_done_callback)(struct fast_task_info *task,
+        const in_addr_64_t client_addr, const bool bInnerPort);
 typedef int (*sf_set_body_length_callback)(struct fast_task_info *task);
-typedef int (*sf_deal_task_func)(struct fast_task_info *task, const int stage);
+typedef char *(*sf_alloc_recv_buffer_callback)(struct fast_task_info *task,
+        const int buff_size, bool *new_alloc);
+typedef int (*sf_deal_task_callback)(struct fast_task_info *task, const int stage);
 typedef int (*sf_recv_timeout_callback)(struct fast_task_info *task);
+typedef int (*sf_send_done_callback)(struct fast_task_info *task,
+        const int length, int *next_stage);
+typedef void (*sf_connect_done_callback)(struct fast_task_info *task,
+        const int err_no);
+
+/* calback for release iovec buffer */
+typedef void (*sf_release_buffer_callback)(struct fast_task_info *task);
+
+typedef int (*sf_error_handler_callback)(const int errnum);
+
+typedef enum {
+    sf_comm_action_continue = 'c',
+    sf_comm_action_break = 'b',
+    sf_comm_action_finish = 'f'
+} SFCommAction;
+
+typedef enum {
+    sf_address_family_auto = 0,
+    sf_address_family_ipv4 = 1,
+    sf_address_family_ipv6 = 2,
+    sf_address_family_both = 3
+} SFAddressFamily;
+
+struct ibv_pd;
+struct sf_listener;
+
+typedef int (*sf_get_connection_size_callback)();
+typedef int (*sf_init_connection_callback)(
+        struct fast_task_info *task, void *arg);
+#define sf_alloc_pd_callback fc_alloc_pd_callback
+
+typedef int (*sf_create_server_callback)(struct sf_listener
+        *listener, int af, const char *bind_addr);
+typedef void (*sf_close_server_callback)(struct sf_listener *listener);
+typedef struct fast_task_info * (*sf_accept_connection_callback)(
+        struct sf_listener *listener);
+typedef int (*sf_async_connect_server_callback)(struct fast_task_info *task);
+typedef int (*sf_async_connect_check_callback)(struct fast_task_info *task);
+typedef void (*sf_close_connection_callback)(struct fast_task_info *task);
+
+typedef ssize_t (*sf_send_data_callback)(struct fast_task_info *task,
+        SFCommAction *action, bool *send_done);
+typedef ssize_t (*sf_recv_data_callback)(struct fast_task_info *task,
+        const bool call_post_recv, SFCommAction *action);
+typedef int (*sf_post_recv_callback)(struct fast_task_info *task);
+
+struct sf_network_handler;
+typedef struct sf_listener {
+    struct sf_network_handler *handler;
+    int port;
+    bool enabled;
+    bool is_inner;
+    union {
+        int sock;  //for socket
+        void *id;  //for rdma_cm
+    };
+    struct sockaddr_in inaddr;  //for accept
+} SFListener;
+
+struct sf_context;
+struct sf_address_family_handler;
+
+typedef struct sf_network_handler {
+    bool enabled;
+    bool explicit_post_recv;
+    FCCommunicationType comm_type;
+    struct sf_address_family_handler *fh;
+    struct ibv_pd *pd;
+
+    SFListener inner;
+    SFListener outer;
+
+    /* for server side */
+    sf_get_connection_size_callback get_connection_size;
+    sf_init_connection_callback init_connection;
+    sf_alloc_pd_callback alloc_pd;
+    sf_create_server_callback create_server;
+    sf_close_server_callback close_server;
+    sf_accept_connection_callback accept_connection;
+
+    /* for client side */
+    sf_async_connect_server_callback async_connect_server;
+    sf_async_connect_check_callback  async_connect_check;
+
+    /* server and client both */
+    sf_close_connection_callback close_connection;
+
+    sf_send_data_callback send_data;
+    sf_recv_data_callback recv_data;
+    sf_post_recv_callback post_recv;  //for rdma
+} SFNetworkHandler;
+
+typedef struct sf_nio_callbacks {
+    TaskCleanUpCallback task_cleanup;
+    sf_deal_task_callback deal_task;
+    sf_set_body_length_callback set_body_length;
+    sf_alloc_recv_buffer_callback alloc_recv_buffer;
+    sf_accept_done_callback accept_done;
+    sf_connect_done_callback connect_done;
+    sf_send_done_callback send_done;
+    sf_recv_timeout_callback task_timeout;
+    sf_release_buffer_callback release_buffer;
+} SFNIOCallbacks;
+
+typedef struct sf_address_family_handler {
+    int af;   //AF_UNSPEC for disabled
+    SFNetworkHandler handlers[SF_NETWORK_HANDLER_COUNT];
+    char inner_bind_addr[IP_ADDRESS_SIZE];
+    char outer_bind_addr[IP_ADDRESS_SIZE];
+    struct sf_context *ctx;
+} SFAddressFamilyHandler;
+
+typedef struct sf_net_buffer_config {
+    int connect_timeout;
+    int network_timeout;
+    int max_connections;
+    int max_pkg_size;
+    int min_buff_size;
+    int max_buff_size;
+} SFNetBufferConfig;
 
 typedef struct sf_context {
+    char name[64];
     struct nio_thread_data *thread_data;
     volatile int thread_count;
-    int outer_sock;
-    int inner_sock;
 
-    int outer_port;
-    int inner_port;
+    bool is_client;    //since v1.2.5
+    bool use_io_uring; //since v1.2.9
+    bool use_send_zc;  //since v1.2.9
+    SFAddressFamily address_family;
+    SFAddressFamilyHandler handlers[SF_ADDRESS_FAMILY_COUNT];
+
+    SFNetBufferConfig net_buffer_cfg;
+
     int accept_threads;
     int work_threads;
 
-    char inner_bind_addr[IP_ADDRESS_SIZE];
-    char outer_bind_addr[IP_ADDRESS_SIZE];
-
     int header_size;
-    bool remove_from_ready_list;
     bool realloc_task_buffer;
-    sf_deal_task_func deal_task;
-    sf_set_body_length_callback set_body_length;
-    sf_accept_done_callback accept_done_func;
-    TaskCleanUpCallback task_cleanup_func;
-    sf_recv_timeout_callback timeout_callback;
+    bool connect_need_log;  //for client connect
+    FCSmartPollingConfig smart_polling;
+
+    SFNIOCallbacks callbacks;
+    struct fast_task_queue free_queue;
 } SFContext;
 
 typedef struct {
     int body_len;      //body length
     short flags;
-    short status;
+    volatile short status;
     unsigned char cmd; //command
 } SFHeaderInfo;
 
@@ -84,16 +222,26 @@ typedef struct {
     SFErrorInfo error;
 } SFResponseInfo;
 
+typedef struct {
+    int64_t req_start_time;  //unit: microsecond (us)
+    SFRequestInfo request;
+    SFResponseInfo response;
+    bool response_done;
+    char log_level;   //level for error log
+    bool need_response;
+} SFCommonTaskContext;
+
 typedef struct sf_binlog_file_position {
     int index;      //current binlog file
     int64_t offset; //current file offset
 } SFBinlogFilePosition;
 
 typedef struct server_binlog_buffer {
-    char *buff;    //the buffer pointer
-    char *current; //for the consumer
-    char *end;     //data end ptr
-    int size;      //the buffer size (capacity)
+    char *buff;       //the buffer pointer
+    char *current;    //for the consumer
+    char *data_end;   //data end ptr
+    char *buff_end;   //buffer end ptr
+    int size;         //the buffer size (capacity)
 } SFBinlogBuffer;
 
 typedef struct sf_space_stat {
@@ -102,9 +250,158 @@ typedef struct sf_space_stat {
     int64_t used;
 } SFSpaceStat;
 
+typedef struct sf_binlog_writer_stat {
+    int64_t total_count;
+    int64_t next_version;
+    int waiting_count;
+    int max_waitings;
+} SFBinlogWriterStat;
+
 typedef struct sf_version_range {
     int64_t first; //including
     int64_t last;  //including
 } SFVersionRange;
 
+typedef struct sf_log_config {
+    int sync_log_buff_interval; //sync log buff to disk every interval seconds
+    bool rotate_everyday;
+    bool compress_old;
+    int compress_days_before;
+    TimeInfo rotate_time;
+    TimeInfo delete_old_time;
+    int keep_days;
+    int64_t rotate_on_size;
+} SFLogConfig;
+
+typedef struct sf_slow_log_config {
+    bool enabled;
+    int log_slower_than_ms;
+    char filename_prefix[64];
+    SFLogConfig log_cfg;
+} SFSlowLogConfig;
+
+typedef struct sf_slow_log_context {
+    SFSlowLogConfig cfg;
+    LogContext ctx;
+} SFSlowLogContext;
+
+typedef enum sf_data_read_rule {
+    sf_data_read_rule_any_available,
+    sf_data_read_rule_slave_first,
+    sf_data_read_rule_master_only
+} SFDataReadRule;
+
+typedef enum sf_net_retry_interval_mode {
+    sf_net_retry_interval_mode_fixed,
+    sf_net_retry_interval_mode_multiple
+} SFNetRetryIntervalMode;
+
+typedef struct sf_net_retry_interval_mode_max_pair {
+    SFNetRetryIntervalMode mode;
+    int max_interval_ms;
+} SFNetRetryIntervalModeMaxPair;
+
+typedef struct sf_net_retry_times_interval_pair {
+    int times;
+    int interval_ms;
+} SFNetRetryTimesIntervalPair;
+
+typedef struct sf_net_retry_config {
+    SFNetRetryIntervalModeMaxPair interval_mm;
+    SFNetRetryTimesIntervalPair connect;
+    SFNetRetryTimesIntervalPair network;
+} SFNetRetryConfig;
+
+typedef struct sf_client_common_config {
+    SFDataReadRule read_rule;  //the rule for read
+    int connect_timeout;
+    int network_timeout;
+    SFNetRetryConfig net_retry_cfg;
+} SFClientCommonConfig;
+
+struct sf_cm_server_entry;
+struct sf_cm_server_ptr_array;
+typedef struct sf_connection_parameters {
+    int buffer_size;
+    struct {
+        struct sf_cm_server_entry *sentry;
+        struct sf_cm_server_ptr_array *old_alives;
+    } cm;  //for connection manager
+    struct idempotency_client_channel *channel;
+} SFConnectionParameters;
+
+typedef struct sf_key_value_array {
+    key_value_pair_t *elts;
+    int count;
+    int alloc;
+} SFKeyValueArray;
+
+typedef struct sf_cmd_option {
+    string_t name;
+    int val;
+    bool has_arg;
+    const char *desc;
+} SFCMDOption;
+
+typedef struct sf_memory_watermark {
+    int64_t low;
+    int64_t high;
+} SFMemoryWatermark;
+
+typedef struct sf_list_limit_info {
+    int offset;
+    int count;
+} SFListLimitInfo;
+
+typedef enum sf_server_group_index_type {
+    sf_server_group_index_type_cluster = 1,
+    sf_server_group_index_type_service
+} SFServerGroupIndexType;
+
+typedef struct sf_cluster_config {
+    FCServerConfig server_cfg;
+    unsigned char md5_digest[SF_CLUSTER_CONFIG_SIGN_LEN];
+    int cluster_group_index;
+    int service_group_index;
+} SFClusterConfig;
+
+typedef struct sf_synchronize_context {
+    pthread_lock_cond_pair_t lcp;
+    int result;
+    union {
+        bool finished;
+        bool ready;
+        int waiting_count;
+    };
+} SFSynchronizeContext;
+
+typedef enum sf_election_quorum {
+    sf_election_quorum_auto,
+    sf_election_quorum_any,
+    sf_election_quorum_majority
+} SFElectionQuorum;
+
+typedef enum sf_replication_quorum {
+    sf_replication_quorum_auto,
+    sf_replication_quorum_any,
+    sf_replication_quorum_majority,
+    sf_replication_quorum_smart
+} SFReplicationQuorum;
+
+typedef struct sf_block_key {
+    int64_t oid;    //object id
+    int64_t offset; //aligned by block size
+    uint64_t hash_code;
+} SFBlockKey;
+
+typedef struct sf_slice_size {
+    int offset;  //offset within the block
+    int length;  //slice length
+} SFSliceSize;
+
+typedef struct sf_block_slice_key_info {
+    SFBlockKey block;
+    SFSliceSize slice;
+} SFBlockSliceKeyInfo;
+
 #endif

src/sf_util.c

@@ -22,7 +22,6 @@
 #include <stdio.h>
 #include <string.h>
 #include <pthread.h>
-#include "sf_global.h"
 #include "sf_define.h"
 #include "sf_util.h"
 
@@ -93,32 +92,175 @@ int sf_printbuffer(char* buffer,int32_t len)
     return(0);
 }
 
-void sf_usage(const char *program)
+void sf_usage_ex(const char *program, const SFCMDOption *other_options)
 {
-    fprintf(stderr, "Usage: %s <config_file> [--without-daemon | --no-daemon] "
-            "[start | stop | restart]\n", program);
-}
+    fprintf(stderr, "\nUsage: %s [options] <config_file> "
+            "[start | stop | restart | status]\n\noptions:\n", program);
 
-void sf_parse_daemon_mode_and_action_ex(int argc, char *argv[],
-        bool *daemon_mode, char **action, const char *default_action)
-{
-    int i;
-
-    *daemon_mode = true;
-    for (i=2; i<argc; i++) {
-        if (strcmp(argv[i], "--without-daemon") == 0 ||
-                strcmp(argv[i], "--no-daemon") == 0)
-        {
-            *daemon_mode = false;
-            break;
+    if (other_options != NULL) {
+        const SFCMDOption *option;
+        option = other_options;
+        while (option->name.str != NULL) {
+            fprintf(stderr, "\t%s\n", option->desc);
+            option++;
         }
     }
 
-    if (argc - (*daemon_mode ? 0 : 1) > 2) {
-        *action = argv[argc - 1];
+    fprintf(stderr, "\t-N | --no-daemon: run in foreground\n"
+            "\t-V | --version: show version info\n"
+            "\t-h | --help: for this usage\n\n");
+}
+
+static int match_option(const char *str, const SFCMDOption *option)
+{
+    const char *start;
+    const char *end;
+
+    if (str[1] == '-') {
+        start = str + 2;
+        while (option->name.str != NULL) {
+            if (strncmp(option->name.str, start,
+                        option->name.len) == 0)
+            {
+                end = start + option->name.len;
+                if (*end == '\0') {
+                    return option->has_arg ? 2 : 1;
+                } else if (*end == '=') {
+                    return 1;
+                }
+            }
+
+            option++;
+        }
+    } else {
+        while (option->name.str != NULL) {
+            if (option->val == str[1]) {
+                if (str[2] == '\0') {
+                    return option->has_arg ? 2 : 1;
+                } else {
+                    return 1;
+                }
+            }
+            option++;
+        }
+    }
+
+    return 0;
+}
+
+
+const char *sf_parse_daemon_mode_and_action_ex(int argc, char *argv[],
+        const Version *version, bool *daemon_mode, char **action,
+        const char *default_action, const SFCMDOption *other_options)
+{
+#define CMD_NORMAL_ARG_COUNT 2
+    int i;
+    int inc;
+    struct {
+        int argc;
+        char *argv[CMD_NORMAL_ARG_COUNT];
+    } normal;
+    const char *config_filepath;
+
+    normal.argc = 0;
+    *daemon_mode = true;
+    i = 1;
+    while (i < argc) {
+        if (argv[i][0] != '-') {
+            if (normal.argc == CMD_NORMAL_ARG_COUNT) {
+                fprintf(stderr, "\nError: too many arguments!\n");
+                sf_usage_ex(argv[0], other_options);
+                return NULL;
+            }
+            normal.argv[normal.argc++] = argv[i++];
+            continue;
+        }
+
+        if (other_options != NULL) {
+            inc = match_option(argv[i], other_options);
+            if (inc > 0) {
+                i += inc;
+                if (i > argc) {
+                    fprintf(stderr, "\nError: expect argument!\n");
+                    sf_usage_ex(argv[0], other_options);
+                    return NULL;
+                }
+                continue;
+            }
+        }
+
+        if (strcmp(argv[i], "-V") == 0 ||
+                strcmp(argv[i], "--version") == 0)
+        {
+            char *last_slash;
+            char *proc_name;
+            if ((last_slash=strrchr(argv[0], '/')) != NULL) {
+                proc_name = last_slash + 1;
+            } else {
+                proc_name = argv[0];
+            }
+            printf("\n%s V%d.%d.%d\n\n", proc_name, version->major,
+                    version->minor, version->patch);
+            return NULL;
+        }
+        if (strcmp(argv[i], "-h") == 0 ||
+                strcmp(argv[i], "--help") == 0)
+        {
+            sf_usage_ex(argv[0], other_options);
+            return NULL;
+        }
+
+        if (strcmp(argv[i], "-N") == 0 ||
+                strcmp(argv[i], "--no-daemon") == 0)
+        {
+            *daemon_mode = false;
+            i++;
+        } else {
+            fprintf(stderr, "\nError: unrecognized option: %s\n", argv[i]);
+            sf_usage_ex(argv[0], other_options);
+            return NULL;
+        }
+    }
+
+    if (normal.argc == 0) {
+        fprintf(stderr, "\nError: expect config file!\n");
+        sf_usage_ex(argv[0], other_options);
+        return NULL;
+    }
+
+    config_filepath = normal.argv[0];
+    if (normal.argc > 1) {
+        *action = normal.argv[1];
     } else {
         *action = (char *)default_action;
     }
+
+    return config_filepath;
+}
+
+void sf_parse_cmd_option_bool(int argc, char *argv[],
+        const string_t *short_option, const string_t *long_option,
+        bool *value)
+{
+    char **pp;
+    char **end;
+    int len;
+
+    *value = false;
+    end = argv + argc;
+    for (pp=argv + 1; pp<end; pp++) {
+        if (**pp != '-') {
+            continue;
+        }
+
+        len = strlen(*pp);
+        if (fc_string_equals2(short_option, *pp, len) ||
+                fc_string_equals2(long_option, *pp, len))
+        {
+            *value = true;
+            break;
+        }
+    }
 }
 
 int sf_logger_init(LogContext *pContext, const char *filename_prefix)
@@ -128,7 +270,7 @@ int sf_logger_init(LogContext *pContext, const char *filename_prefix)
         return result;
     }
 
-    if ((result=log_set_prefix_ex(pContext, g_sf_global_vars.base_path,
+    if ((result=log_set_prefix_ex(pContext, SF_G_BASE_PATH_STR,
                     filename_prefix)) != 0)
     {
         return result;
@@ -139,36 +281,27 @@ int sf_logger_init(LogContext *pContext, const char *filename_prefix)
     return 0;
 }
 
-ScheduleEntry *sf_logger_set_schedule_entry(struct log_context *pContext,
-        ScheduleEntry *pScheduleEntry)
+ScheduleEntry *sf_logger_set_schedule_entries(struct log_context *pContext,
+        SFLogConfig *log_cfg, ScheduleEntry *pScheduleEntry)
 {
-    pScheduleEntry->id = sched_generate_next_id();
-    pScheduleEntry->time_base.hour = TIME_NONE;
-    pScheduleEntry->time_base.minute = TIME_NONE;
-    pScheduleEntry->interval = g_sf_global_vars.sync_log_buff_interval;
-    pScheduleEntry->task_func = log_sync_func;
-    pScheduleEntry->func_args = pContext;
+    INIT_SCHEDULE_ENTRY(*pScheduleEntry, sched_generate_next_id(),
+            TIME_NONE, TIME_NONE, 0, log_cfg->sync_log_buff_interval,
+            log_sync_func, pContext);
     pScheduleEntry++;
 
-    pScheduleEntry->id = sched_generate_next_id();
-    pScheduleEntry->time_base.hour = 0;
-    pScheduleEntry->time_base.minute = 0;
-    pScheduleEntry->time_base.second = 0;
-    pScheduleEntry->interval = 86400;
-    pScheduleEntry->task_func = log_notify_rotate;
-    pScheduleEntry->func_args = pContext;
-    pScheduleEntry++;
+    if (log_cfg->rotate_everyday) {
+        INIT_SCHEDULE_ENTRY_EX(*pScheduleEntry, sched_generate_next_id(),
+                log_cfg->rotate_time, 86400, log_notify_rotate, pContext);
+        pScheduleEntry++;
+    }
 
-    if (g_sf_global_vars.log_file_keep_days > 0) {
-        log_set_keep_days(pContext, g_sf_global_vars.log_file_keep_days);
-
-        pScheduleEntry->id = sched_generate_next_id();
-        pScheduleEntry->time_base.hour = 1;
-        pScheduleEntry->time_base.minute = 30;
-        pScheduleEntry->time_base.second = 0;
-        pScheduleEntry->interval = 86400;
-        pScheduleEntry->task_func = log_delete_old_files;
-        pScheduleEntry->func_args = pContext;
+    if ((log_cfg->rotate_everyday || log_cfg->rotate_on_size > 0) &&
+            (log_cfg->keep_days > 0))
+    {
+        log_set_keep_days(pContext, log_cfg->keep_days);
+        INIT_SCHEDULE_ENTRY_EX(*pScheduleEntry, sched_generate_next_id(),
+                log_cfg->delete_old_time, 86400, log_delete_old_files,
+                pContext);
         pScheduleEntry++;
     }
 
@@ -178,6 +311,8 @@ ScheduleEntry *sf_logger_set_schedule_entry(struct log_context *pContext,
 const char *sf_strerror(const int errnum)
 {
     switch (errnum) {
+        case SF_CLUSTER_ERROR_BINLOG_MISSED:
+            return "binlog missed";
         case SF_CLUSTER_ERROR_BINLOG_INCONSISTENT:
             return "binlog inconsistent";
         case SF_CLUSTER_ERROR_LEADER_INCONSISTENT:
@@ -192,6 +327,16 @@ const char *sf_strerror(const int errnum)
             return "idempotency channel not exist";
         case SF_RETRIABLE_ERROR_CHANNEL_INVALID:
             return "idempotency channel is invalid";
+        case SF_ERROR_EINVAL:
+            return STRERROR(EINVAL);
+        case SF_ERROR_EAGAIN:
+            return STRERROR(EAGAIN);
+        case SF_ERROR_EINPROGRESS:
+            return STRERROR(EINPROGRESS);
+        case SF_ERROR_EOVERFLOW:
+            return STRERROR(EOVERFLOW);
+        case SF_ERROR_ENODATA:
+            return STRERROR(ENODATA);
         default:
             return STRERROR(errnum);
     }

src/sf_util.h

@@ -18,8 +18,12 @@
 #ifndef _SF_UTIL_H_
 #define _SF_UTIL_H_
 
+#include <getopt.h>
 #include "fastcommon/logger.h"
 #include "fastcommon/sched_thread.h"
+#include "sf_define.h"
+#include "sf_types.h"
+#include "sf_global.h"
 
 #ifdef DEBUG_FLAG  /*only for format check*/
 
@@ -56,28 +60,115 @@ __FILE__, eln, eres, emsg, strerror(eres))
 
 #define dszoffset(cls, mem) ((char*)&((cls*)0)->mem -  ((char*)0))
 
-#define sf_parse_daemon_mode_and_action(argc, argv, daemon_mode, action) \
-    sf_parse_daemon_mode_and_action_ex(argc, argv, daemon_mode, action, "start")
+#define sf_usage(program) sf_usage_ex(program, NULL)
+
+#define sf_parse_daemon_mode_and_action(argc, argv, \
+        version, daemon_mode, action) \
+    sf_parse_daemon_mode_and_action_ex(argc, argv, \
+            version, daemon_mode, action, "start", NULL)
+
+#define SF_COMMON_OPT_STRING  "NVh"
+#define SF_COMMON_LONG_OPTIONS  \
+    {"no-daemon", no_argument, NULL, 'N'}, \
+    {"version",   no_argument, NULL, 'V'},   \
+    {"help",      no_argument, NULL, 'h'}
 
 #ifdef __cplusplus
 extern "C" {
 #endif
 
-int64_t getticks() ;
+int64_t getticks();
 
-void log_plus(const int priority, const char* file, int line, const char* fmt, ...);
+void log_plus(const int priority, const char *file,
+        int line, const char *fmt, ...);
 
-int sf_printbuffer(char* buffer,int32_t len);
+int sf_printbuffer(char *buffer,int32_t len);
 
-void sf_usage(const char *program);
+void sf_usage_ex(const char *program, const SFCMDOption *other_options);
 
-void sf_parse_daemon_mode_and_action_ex(int argc, char *argv[],
-        bool *daemon_mode, char **action, const char *default_action);
+const char *sf_parse_daemon_mode_and_action_ex(int argc, char *argv[],
+        const Version *version, bool *daemon_mode, char **action,
+        const char *default_action, const SFCMDOption *other_options);
+
+void sf_parse_cmd_option_bool(int argc, char *argv[],
+        const string_t *short_option, const string_t *long_option,
+        bool *value);
 
 int sf_logger_init(LogContext *pContext, const char *filename_prefix);
 
-ScheduleEntry *sf_logger_set_schedule_entry(struct log_context *pContext,
-        ScheduleEntry *pScheduleEntry);
+ScheduleEntry *sf_logger_set_schedule_entries(struct log_context *pContext,
+        SFLogConfig *log_cfg, ScheduleEntry *pScheduleEntry);
+
+static inline void sf_logger_setup_schedule(struct log_context *pContext,
+        SFLogConfig *log_cfg, ScheduleArray *scheduleArray)
+{
+    ScheduleEntry *scheduleEntry;
+    scheduleEntry = sf_logger_set_schedule_entries(pContext,
+            log_cfg, scheduleArray->entries);
+    scheduleArray->count = scheduleEntry - scheduleArray->entries;
+}
+
+static inline int sf_unify_errno(const int errnum)
+{
+    switch (errnum) {
+        case EBUSY:
+            return SF_ERROR_EBUSY;
+        case EINVAL:
+            return SF_ERROR_EINVAL;
+        case EAGAIN:
+            return SF_ERROR_EAGAIN;
+        case EINPROGRESS:
+            return SF_ERROR_EINPROGRESS;
+        case EOVERFLOW:
+            return SF_ERROR_EOVERFLOW;
+        case EOPNOTSUPP:
+            return SF_ERROR_EOPNOTSUPP;
+        case ENODATA:
+            return SF_ERROR_ENODATA;
+        case ENOLINK:
+            return SF_ERROR_ENOLINK;
+        case ENOTEMPTY:
+            return SF_ERROR_ENOTEMPTY;
+        case ELOOP:
+            return SF_ERROR_ELOOP;
+        default:
+            return errnum;
+    }
+}
+
+static inline int sf_localize_errno(int errnum)
+{
+    if (SF_G_ERROR_HANDLER != NULL) {
+        errnum = SF_G_ERROR_HANDLER(errnum);
+    }
+
+    switch (errnum) {
+        case SF_ERROR_EBUSY:
+            return EBUSY;
+        case SF_ERROR_EINVAL:
+            return EINVAL;
+        case SF_ERROR_EAGAIN:
+            return EAGAIN;
+        case SF_ERROR_EINPROGRESS:
+            return EINPROGRESS;
+        case SF_ERROR_EOVERFLOW:
+            return EOVERFLOW;
+        case SF_ERROR_EOPNOTSUPP:
+            return EOPNOTSUPP;
+        case SF_ERROR_ENODATA:
+            return ENODATA;
+        case SF_SESSION_ERROR_NOT_EXIST:
+            return EPERM;
+        case SF_ERROR_ENOLINK:
+            return ENOLINK;
+        case SF_ERROR_ENOTEMPTY:
+            return ENOTEMPTY;
+        case SF_ERROR_ELOOP:
+            return ELOOP;
+        default:
+            return errnum;
+    }
+}
 
 const char *sf_strerror(const int errnum);